Datasets:
microtensor-io
/
agent-sessions

Dataset card Data Studio Files
xet
 Community
item_id
stringlengths
21
26
family
stringclasses
383 values
twin_of
stringlengths
0
21
system
stringlengths
1.51k
1.51k
user
stringlengths
18.6k
288k
input_tokens
int64
6.38k
123k
budget_tokens
int64
1k
3.07k
cut
int64
19
151
steps
int64
32
166
resolved
bool
1 class
image
stringlengths
25
33
source
stringclasses
1 value
facts
stringlengths
585
1.33k
probes
stringlengths
1.29k
1.44k
ctx_compress-s1-01457
tokenforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope_safe raises ValueError when a value itself contains an equals sign. It lives in src/tokenforge/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tokenforge layou...
42,118
2,105
90
99
false
synth:2386905581134506283@1
synth
[{"aliases": ["src/tokenforge/io.py", "io.py"], "kind": "file", "value": "src/tokenforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/tokenforge/audit.py", "audit.py"], "kind": "file", "value": "src/tokenforge/audit.py"}, {"aliases": ["data/tok...
[{"answers": ["src/tokenforge/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-01457-twin
tokenforge
ctx_compress-s1-01457
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope_safe raises ValueError when a value itself contains an equals sign. It lives in src/tokenforge/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tokenforge layou...
24,080
1,204
58
99
false
synth:2386905581134506283@1
synth
[{"aliases": ["src/tokenforge/io.py", "io.py"], "kind": "file", "value": "src/tokenforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/tokenforge/audit.py", "audit.py"], "kind": "file", "value": "src/tokenforge/audit.py"}, {"aliases": ["data/tok...
[{"answers": ["src/tokenforge/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-01458
stockworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] pick_pallet double counts carried amounts when the period closes on a weekend. It lives in src/stockworks/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockworks lay...
77,698
3,072
98
119
false
synth:1125129970592108431@2.22285
synth
[{"aliases": ["src/stockworks/core.py", "core.py"], "kind": "file", "value": "src/stockworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/stockworks/format.py", "format.py"], "kind": "file", "value": "src/stockworks/format.py"}, {"alias...
[{"answers": ["src/stockworks/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01458-twin
stockworks
ctx_compress-s1-01458
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] pick_pallet double counts carried amounts when the period closes on a weekend. It lives in src/stockworks/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockworks lay...
48,667
2,433
68
119
false
synth:1125129970592108431@2.22285
synth
[{"aliases": ["src/stockworks/core.py", "core.py"], "kind": "file", "value": "src/stockworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/stockworks/format.py", "format.py"], "kind": "file", "value": "src/stockworks/format.py"}, {"alias...
[{"answers": ["src/stockworks/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01459
policykit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_token skips the first page; page 1 returns what should be page 2. It lives in src/policykit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the policykit layout. [to...
43,594
2,179
83
100
false
synth:5845961432896266715@1
synth
[{"aliases": ["src/policykit/core.py", "core.py"], "kind": "file", "value": "src/policykit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/policy_4.csv", "policy_4.csv"], "kind": "file", "value": "data/policy_4.csv"}, {"aliases": ["data/c...
[{"answers": ["src/policykit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01459-twin
policykit
ctx_compress-s1-01459
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_token skips the first page; page 1 returns what should be page 2. It lives in src/policykit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the policykit layout. [to...
21,433
1,071
45
100
false
synth:5845961432896266715@1
synth
[{"aliases": ["src/policykit/core.py", "core.py"], "kind": "file", "value": "src/policykit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/policy_4.csv", "policy_4.csv"], "kind": "file", "value": "data/policy_4.csv"}, {"aliases": ["data/c...
[{"answers": ["src/policykit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most r...
ctx_compress-s1-01460
cadencekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_booking returns one item too few on every page. It lives in src/cadencekit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cadencekit layout. [tool_use kind=...
75,102
3,072
98
113
false
synth:2783481452114294225@1.8263
synth
[{"aliases": ["src/cadencekit/limits.py", "limits.py"], "kind": "file", "value": "src/cadencekit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/shift_6.csv", "shift_6.csv"], "kind": "file", "value": "data/shift_6.csv"}, {"aliases...
[{"answers": ["src/cadencekit/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01460-twin
cadencekit
ctx_compress-s1-01460
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_booking returns one item too few on every page. It lives in src/cadencekit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cadencekit layout. [tool_use kind=...
48,255
2,412
64
113
false
synth:2783481452114294225@1.8263
synth
[{"aliases": ["src/cadencekit/limits.py", "limits.py"], "kind": "file", "value": "src/cadencekit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/shift_6.csv", "shift_6.csv"], "kind": "file", "value": "data/shift_6.csv"}, {"aliases...
[{"answers": ["src/cadencekit/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01461
rosterdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_cadence skips the first page; page 1 returns what should be page 2. It lives in src/rosterdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the rosterdesk lay...
46,624
2,331
76
87
false
synth:5116587215226682275@1
synth
[{"aliases": ["src/rosterdesk/format.py", "format.py"], "kind": "file", "value": "src/rosterdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/slot_1.csv", "slot_1.csv"], "kind": "file", "value": "data/slot_1.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/rosterdesk/format.py...
ctx_compress-s1-01461-twin
rosterdesk
ctx_compress-s1-01461
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_cadence skips the first page; page 1 returns what should be page 2. It lives in src/rosterdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the rosterdesk lay...
24,448
1,222
46
87
false
synth:5116587215226682275@1
synth
[{"aliases": ["src/rosterdesk/format.py", "format.py"], "kind": "file", "value": "src/rosterdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/slot_1.csv", "slot_1.csv"], "kind": "file", "value": "data/slot_1.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/rosterdesk/store.py", "store.py"]...
ctx_compress-s1-01462
bindesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_bin_page raises KeyError for a missing key instead of returning the default. It lives in src/bindesk/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the bindesk...
105,732
3,072
136
147
false
synth:7146815239315620802@1.86223
synth
[{"aliases": ["src/bindesk/report.py", "report.py"], "kind": "file", "value": "src/bindesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/warehouse_9.csv", "warehouse_9.csv"], "kind": "file", "value": "data/warehouse_9.csv"}, {"a...
[{"answers": ["src/bindesk/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01462-twin
bindesk
ctx_compress-s1-01462
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_bin_page raises KeyError for a missing key instead of returning the default. It lives in src/bindesk/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the bindesk...
45,144
2,257
70
147
false
synth:7146815239315620802@1.86223
synth
[{"aliases": ["src/bindesk/report.py", "report.py"], "kind": "file", "value": "src/bindesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/warehouse_9.csv", "warehouse_9.csv"], "kind": "file", "value": "data/warehouse_9.csv"}, {"a...
[{"answers": ["src/bindesk/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01463
scopedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_scope lets defaults overwrite explicit overrides. It lives in src/scopedesk/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the scopedesk layout. [tool_use kind=search n...
108,606
3,072
102
123
false
synth:7014439600336242636@2.239
synth
[{"aliases": ["src/scopedesk/io.py", "io.py"], "kind": "file", "value": "src/scopedesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_4.csv", "grant_4.csv"], "kind": "file", "value": "data/grant_4.csv"}, {"aliases": ["docs/changelog.md", "...
[{"answers": ["src/scopedesk/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recen...
ctx_compress-s1-01463-twin
scopedesk
ctx_compress-s1-01463
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_scope lets defaults overwrite explicit overrides. It lives in src/scopedesk/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the scopedesk layout. [tool_use kind=search n...
62,881
3,072
64
123
false
synth:7014439600336242636@2.239
synth
[{"aliases": ["src/scopedesk/io.py", "io.py"], "kind": "file", "value": "src/scopedesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_4.csv", "grant_4.csv"], "kind": "file", "value": "data/grant_4.csv"}, {"aliases": ["docs/changelog.md", "...
[{"answers": ["src/scopedesk/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most recen...
ctx_compress-s1-01464
stockflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_sku double counts carried amounts when the period closes on a weekend. It lives in src/stockflow/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockflow l...
75,273
3,072
103
114
false
synth:5338791472084455981@2.11128
synth
[{"aliases": ["src/stockflow/format.py", "format.py"], "kind": "file", "value": "src/stockflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/pallet_2.csv", "pallet_2.csv"], "kind": "file", "value": "data/pallet_2.csv"}, {"aliase...
[{"answers": ["src/stockflow/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01464-twin
stockflow
ctx_compress-s1-01464
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_sku double counts carried amounts when the period closes on a weekend. It lives in src/stockflow/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockflow l...
44,890
2,244
65
114
false
synth:5338791472084455981@2.11128
synth
[{"aliases": ["src/stockflow/format.py", "format.py"], "kind": "file", "value": "src/stockflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/pallet_2.csv", "pallet_2.csv"], "kind": "file", "value": "data/pallet_2.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/sku_3.csv", "sku_3.csv"], "kind"...
ctx_compress-s1-01465
traceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_counter skips the first page; page 1 returns what should be page 2. It lives in src/traceledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the traceledger lay...
66,067
3,072
101
118
false
synth:5483534223261239187@1.62827
synth
[{"aliases": ["src/traceledger/format.py", "format.py"], "kind": "file", "value": "src/traceledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/counter_1.csv", "counter_1.csv"], "kind": "file", "value": "data/counter_1.csv"}, {...
[{"answers": ["src/traceledger/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01465-twin
traceledger
ctx_compress-s1-01465
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_counter skips the first page; page 1 returns what should be page 2. It lives in src/traceledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the traceledger lay...
36,272
1,813
57
118
false
synth:5483534223261239187@1.62827
synth
[{"aliases": ["src/traceledger/format.py", "format.py"], "kind": "file", "value": "src/traceledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/counter_1.csv", "counter_1.csv"], "kind": "file", "value": "data/counter_1.csv"}, {...
[{"answers": ["src/traceledger/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01466
binledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_pallet double counts carried amounts when the period closes on a weekend. It lives in src/binledger/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the binledger layout. ...
72,069
3,072
87
98
false
synth:1806515005776813056@1.7914
synth
[{"aliases": ["src/binledger/io.py", "io.py"], "kind": "file", "value": "src/binledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/binledger/tables.py", "tables.py"], "kind": "file", "value": "src/binledger/tables.py"}, {"aliases": ["src/binle...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["io.py", "src/binledger/io.py"], "kind...
ctx_compress-s1-01466-twin
binledger
ctx_compress-s1-01466
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_pallet double counts carried amounts when the period closes on a weekend. It lives in src/binledger/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the binledger layout. ...
47,367
2,368
57
98
false
synth:1806515005776813056@1.7914
synth
[{"aliases": ["src/binledger/io.py", "io.py"], "kind": "file", "value": "src/binledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/binledger/tables.py", "tables.py"], "kind": "file", "value": "src/binledger/tables.py"}, {"aliases": ["src/binle...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["schema.py", "src/binledger/schema.py"...
ctx_compress-s1-01467
chunkkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_chunk raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/chunkkit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the chunkki...
89,613
3,072
97
118
false
synth:106747047914356766@1.57515
synth
[{"aliases": ["src/chunkkit/core.py", "core.py"], "kind": "file", "value": "src/chunkkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/chunk_3.csv", "chunk_3.csv"], "kind": "file", "value": "data/chunk_3.csv"}, {"aliases": ["src/chunkki...
[{"answers": ["src/chunkkit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-01467-twin
chunkkit
ctx_compress-s1-01467
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_chunk raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/chunkkit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the chunkki...
49,694
2,484
63
118
false
synth:106747047914356766@1.57515
synth
[{"aliases": ["src/chunkkit/core.py", "core.py"], "kind": "file", "value": "src/chunkkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/chunk_3.csv", "chunk_3.csv"], "kind": "file", "value": "data/chunk_3.csv"}, {"aliases": ["src/chunkki...
[{"answers": ["src/chunkkit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-01468
balanceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reconcile_receipt leaves a trailing dash when the title ends in punctuation. It lives in src/balanceledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the balancele...
91,915
3,072
99
126
false
synth:8521826144001531951@2.0185
synth
[{"aliases": ["src/balanceledger/report.py", "report.py"], "kind": "file", "value": "src/balanceledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/payout_4.csv", "payout_4.csv"], "kind": "file", "value": "data/payout_4.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/tariff_6.csv", "tariff_6.csv"], ...
ctx_compress-s1-01468-twin
balanceledger
ctx_compress-s1-01468
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reconcile_receipt leaves a trailing dash when the title ends in punctuation. It lives in src/balanceledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the balancele...
56,422
2,821
73
126
false
synth:8521826144001531951@2.0185
synth
[{"aliases": ["src/balanceledger/report.py", "report.py"], "kind": "file", "value": "src/balanceledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/payout_4.csv", "payout_4.csv"], "kind": "file", "value": "data/payout_4.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/tariff_9.csv", "tariff_9.csv"], ...
ctx_compress-s1-01469
legkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_beacon_batch ignores the cap, so late attempts wait for minutes. It lives in src/legkit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the legkit layout. [tool_use...
27,970
1,398
50
63
false
synth:6088893318698279503@1
synth
[{"aliases": ["src/legkit/core.py", "core.py"], "kind": "file", "value": "src/legkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/legkit/rules.py", "rules.py"], "kind": "file", "value": "src/legkit/rules.py"}, {"aliases": ["docs/changel...
[{"answers": ["src/legkit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-01469-twin
legkit
ctx_compress-s1-01469
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_beacon_batch ignores the cap, so late attempts wait for minutes. It lives in src/legkit/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the legkit layout. [tool_use...
15,184
1,000
34
63
false
synth:6088893318698279503@1
synth
[{"aliases": ["src/legkit/core.py", "core.py"], "kind": "file", "value": "src/legkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/legkit/rules.py", "rules.py"], "kind": "file", "value": "src/legkit/rules.py"}, {"aliases": ["docs/changel...
[{"answers": ["src/legkit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-01470
warehouseforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_stock_window averages one value too many once the window is full. It lives in src/warehouseforge/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the warehouseforge la...
96,679
3,072
110
123
false
synth:9133085788312509539@1.69918
synth
[{"aliases": ["src/warehouseforge/cli.py", "cli.py"], "kind": "file", "value": "src/warehouseforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/warehouseforge/format.py", "format.py"], "kind": "file", "value": "src/warehouseforge/format.py"...
[{"answers": ["src/warehouseforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01470-twin
warehouseforge
ctx_compress-s1-01470
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_stock_window averages one value too many once the window is full. It lives in src/warehouseforge/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the warehouseforge la...
45,709
2,285
64
123
false
synth:9133085788312509539@1.69918
synth
[{"aliases": ["src/warehouseforge/cli.py", "cli.py"], "kind": "file", "value": "src/warehouseforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/warehouseforge/format.py", "format.py"], "kind": "file", "value": "src/warehouseforge/format.py"...
[{"answers": ["src/warehouseforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01471
reorderkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_reorder skips the first page; page 1 returns what should be page 2. It lives in src/reorderkit/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the reorderkit lay...
27,913
1,395
50
63
false
synth:2267103120590602008@1
synth
[{"aliases": ["src/reorderkit/rules.py", "rules.py"], "kind": "file", "value": "src/reorderkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/reorderkit/cli.py", "cli.py"], "kind": "file", "value": "src/reorderkit/cli.py"}, {"aliases"...
[{"answers": ["src/reorderkit/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01471-twin
reorderkit
ctx_compress-s1-01471
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_reorder skips the first page; page 1 returns what should be page 2. It lives in src/reorderkit/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the reorderkit lay...
19,820
1,000
34
63
false
synth:2267103120590602008@1
synth
[{"aliases": ["src/reorderkit/rules.py", "rules.py"], "kind": "file", "value": "src/reorderkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/reorderkit/cli.py", "cli.py"], "kind": "file", "value": "src/reorderkit/cli.py"}, {"aliases"...
[{"answers": ["src/reorderkit/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01472
spanflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_series misses keys that differ only in case. It lives in src/spanflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the spanflow layout. [tool_use kind=search name=Bash]...
27,009
1,350
57
68
false
synth:3493558403044896585@1
synth
[{"aliases": ["src/spanflow/io.py", "io.py"], "kind": "file", "value": "src/spanflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/spanflow/schema.py", "schema.py"], "kind": "file", "value": "src/spanflow/schema.py"}, {"aliases": ["src/spanflow/...
[{"answers": ["src/spanflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent...
ctx_compress-s1-01472-twin
spanflow
ctx_compress-s1-01472
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_series misses keys that differ only in case. It lives in src/spanflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the spanflow layout. [tool_use kind=search name=Bash]...
13,362
1,000
31
68
false
synth:3493558403044896585@1
synth
[{"aliases": ["src/spanflow/io.py", "io.py"], "kind": "file", "value": "src/spanflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/spanflow/schema.py", "schema.py"], "kind": "file", "value": "src/spanflow/schema.py"}, {"aliases": ["src/spanflow/...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["io.py", "src/spanflow/io.py"], "kind"...
ctx_compress-s1-01473
probeledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_gauge_window double counts carried amounts when the period closes on a weekend. It lives in src/probeledger/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the probe...
60,935
3,046
106
119
false
synth:8564346569722989706@1.08155
synth
[{"aliases": ["src/probeledger/rules.py", "rules.py"], "kind": "file", "value": "src/probeledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/span_4.csv", "span_4.csv"], "kind": "file", "value": "data/span_4.csv"}, {"aliases": ["da...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/sample_3.csv", "sample_3.csv"], ...
ctx_compress-s1-01473-twin
probeledger
ctx_compress-s1-01473
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_gauge_window double counts carried amounts when the period closes on a weekend. It lives in src/probeledger/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the probe...
46,478
2,323
66
119
false
synth:8564346569722989706@1.08155
synth
[{"aliases": ["src/probeledger/rules.py", "rules.py"], "kind": "file", "value": "src/probeledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/span_4.csv", "span_4.csv"], "kind": "file", "value": "data/span_4.csv"}, {"aliases": ["da...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["limits.py", "src/probeledger/limits.p...
ctx_compress-s1-01474
slotforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_window_safe raises ShiftShapeError for bytes input in strict mode instead of decoding it. It lives in src/slotforge/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mappin...
44,442
2,222
61
76
false
synth:8615510460198394624@1
synth
[{"aliases": ["src/slotforge/store.py", "store.py"], "kind": "file", "value": "src/slotforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/quota_0.csv", "quota_0.csv"], "kind": "file", "value": "data/quota_0.csv"}, {"aliases": ["doc...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/shift_2.csv", "shift_2.csv"], "k...
ctx_compress-s1-01474-twin
slotforge
ctx_compress-s1-01474
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_window_safe raises ShiftShapeError for bytes input in strict mode instead of decoding it. It lives in src/slotforge/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mappin...
31,331
1,566
41
76
false
synth:8615510460198394624@1
synth
[{"aliases": ["src/slotforge/store.py", "store.py"], "kind": "file", "value": "src/slotforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/quota_0.csv", "quota_0.csv"], "kind": "file", "value": "data/quota_0.csv"}, {"aliases": ["doc...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["cli.py", "src/slotforge/cli.py"], "ki...
ctx_compress-s1-01475
tileledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_fix returns the upper bound for every value instead of clamping into the range. It lives in src/tileledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
97,644
3,072
125
136
false
synth:8449536570511598455@2.02493
synth
[{"aliases": ["src/tileledger/report.py", "report.py"], "kind": "file", "value": "src/tileledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/tileledger/limits.py", "limits.py"], "kind": "file", "value": "src/tileledger/limits.p...
[{"answers": ["src/tileledger/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01475-twin
tileledger
ctx_compress-s1-01475
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_fix returns the upper bound for every value instead of clamping into the range. It lives in src/tileledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
51,590
2,579
67
136
false
synth:8449536570511598455@2.02493
synth
[{"aliases": ["src/tileledger/report.py", "report.py"], "kind": "file", "value": "src/tileledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/tileledger/limits.py", "limits.py"], "kind": "file", "value": "src/tileledger/limits.p...
[{"answers": ["src/tileledger/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01476
cuedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_slug_strict produces doubled dashes for consecutive separators. It lives in src/cuedesk/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cuedesk layout. [tool_use k...
82,817
3,072
102
109
false
synth:6787128310600516012@2.029
synth
[{"aliases": ["src/cuedesk/core.py", "core.py"], "kind": "file", "value": "src/cuedesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/cuedesk/io.py", "io.py"], "kind": "file", "value": "src/cuedesk/io.py"}, {"aliases": ["src/cuedesk/schem...
[{"answers": ["src/cuedesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01476-twin
cuedesk
ctx_compress-s1-01476
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_slug_strict produces doubled dashes for consecutive separators. It lives in src/cuedesk/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cuedesk layout. [tool_use k...
46,245
2,312
58
109
false
synth:6787128310600516012@2.029
synth
[{"aliases": ["src/cuedesk/core.py", "core.py"], "kind": "file", "value": "src/cuedesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/cuedesk/io.py", "io.py"], "kind": "file", "value": "src/cuedesk/io.py"}, {"aliases": ["src/cuedesk/schem...
[{"answers": ["src/cuedesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01477
clausedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/clausedesk/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cl...
21,447
1,072
59
66
false
synth:8655250498601056500@1
synth
[{"aliases": ["src/clausedesk/compat.py", "compat.py"], "kind": "file", "value": "src/clausedesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/clausedesk/store.py", "store.py"], "kind": "file", "value": "src/clausedesk/store.py"}...
[{"answers": ["src/clausedesk/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01477-twin
clausedesk
ctx_compress-s1-01477
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/clausedesk/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cl...
10,546
1,000
33
66
false
synth:8655250498601056500@1
synth
[{"aliases": ["src/clausedesk/compat.py", "compat.py"], "kind": "file", "value": "src/clausedesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/clausedesk/store.py", "store.py"], "kind": "file", "value": "src/clausedesk/store.py"}...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01478
creditledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_credit averages one value too many once the window is full. It lives in src/creditledger/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the creditledger layout. [too...
86,445
3,072
113
124
false
synth:410459454757951581@2.30808
synth
[{"aliases": ["src/creditledger/store.py", "store.py"], "kind": "file", "value": "src/creditledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/invoice_5.csv", "invoice_5.csv"], "kind": "file", "value": "data/invoice_5.csv"}, {"ali...
[{"answers": ["src/creditledger/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01478-twin
creditledger
ctx_compress-s1-01478
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_credit averages one value too many once the window is full. It lives in src/creditledger/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the creditledger layout. [too...
42,675
2,133
57
124
false
synth:410459454757951581@2.30808
synth
[{"aliases": ["src/creditledger/store.py", "store.py"], "kind": "file", "value": "src/creditledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/invoice_5.csv", "invoice_5.csv"], "kind": "file", "value": "data/invoice_5.csv"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/invoice_5.csv", "invoice_5.csv"]...
ctx_compress-s1-01479
waypointkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] bound_beacon_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/waypointkit/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the waypoin...
57,448
2,872
95
106
false
synth:2829423526195358600@1.84977
synth
[{"aliases": ["src/waypointkit/io.py", "io.py"], "kind": "file", "value": "src/waypointkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/waypointkit/schema.py", "schema.py"], "kind": "file", "value": "src/waypointkit/schema.py"}, {"aliases": ["d...
[{"answers": ["src/waypointkit/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01479-twin
waypointkit
ctx_compress-s1-01479
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] bound_beacon_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/waypointkit/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the waypoin...
31,029
1,551
49
106
false
synth:2829423526195358600@1.84977
synth
[{"aliases": ["src/waypointkit/io.py", "io.py"], "kind": "file", "value": "src/waypointkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/waypointkit/schema.py", "schema.py"], "kind": "file", "value": "src/waypointkit/schema.py"}, {"aliases": ["d...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/fix_3.csv", "fix_3.csv"], "kind"...
ctx_compress-s1-01480
payoutledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_refund raises ValueError when a value itself contains an equals sign. It lives in src/payoutledger/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the payoutledger layout....
66,712
3,072
97
108
false
synth:7590600663778552872@1.8743
synth
[{"aliases": ["src/payoutledger/io.py", "io.py"], "kind": "file", "value": "src/payoutledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/payoutledger/store.py", "store.py"], "kind": "file", "value": "src/payoutledger/store.py"}, {"aliases": ["...
[{"answers": ["src/payoutledger/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most re...
ctx_compress-s1-01480-twin
payoutledger
ctx_compress-s1-01480
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_refund raises ValueError when a value itself contains an equals sign. It lives in src/payoutledger/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the payoutledger layout....
33,734
1,686
51
108
false
synth:7590600663778552872@1.8743
synth
[{"aliases": ["src/payoutledger/io.py", "io.py"], "kind": "file", "value": "src/payoutledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/payoutledger/store.py", "store.py"], "kind": "file", "value": "src/payoutledger/store.py"}, {"aliases": ["...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_store.py", "tests/test_store.py"...
ctx_compress-s1-01481
rosterflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_roster averages one value too many once the window is full. It lives in src/rosterflow/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the rosterflow layout. [tool...
25,818
1,290
50
59
false
synth:1670112441972730253@1
synth
[{"aliases": ["src/rosterflow/audit.py", "audit.py"], "kind": "file", "value": "src/rosterflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/calendar_2.csv", "calendar_2.csv"], "kind": "file", "value": "data/calendar_2.csv"}, {"alia...
[{"answers": ["src/rosterflow/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01481-twin
rosterflow
ctx_compress-s1-01481
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_roster averages one value too many once the window is full. It lives in src/rosterflow/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the rosterflow layout. [tool...
11,857
1,000
27
59
false
synth:1670112441972730253@1
synth
[{"aliases": ["src/rosterflow/audit.py", "audit.py"], "kind": "file", "value": "src/rosterflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/calendar_2.csv", "calendar_2.csv"], "kind": "file", "value": "data/calendar_2.csv"}, {"alia...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["audit.py", "src/rosterflow/audit.py"]...
ctx_compress-s1-01482
binflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_shipment misses keys that differ only in case. It lives in src/binflow/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the binflow layout. [tool_use kind=search nam...
44,614
2,230
79
92
false
synth:3190625842000516801@1.2052
synth
[{"aliases": ["src/binflow/compat.py", "compat.py"], "kind": "file", "value": "src/binflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ...
[{"answers": ["src/binflow/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01482-twin
binflow
ctx_compress-s1-01482
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_shipment misses keys that differ only in case. It lives in src/binflow/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the binflow layout. [tool_use kind=search nam...
24,692
1,234
43
92
false
synth:3190625842000516801@1.2052
synth
[{"aliases": ["src/binflow/compat.py", "compat.py"], "kind": "file", "value": "src/binflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01483
claimflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_grant_window double counts carried amounts when the period closes on a weekend. It lives in src/claimflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cla...
25,562
1,278
49
56
false
synth:530168814985809198@1
synth
[{"aliases": ["src/claimflow/report.py", "report.py"], "kind": "file", "value": "src/claimflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/claimflow/compat.py", "compat.py"], "kind": "file", "value": "src/claimflow/compat.py"},...
[{"answers": ["src/claimflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01483-twin
claimflow
ctx_compress-s1-01483
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_grant_window double counts carried amounts when the period closes on a weekend. It lives in src/claimflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cla...
11,747
1,000
27
56
false
synth:530168814985809198@1
synth
[{"aliases": ["src/claimflow/report.py", "report.py"], "kind": "file", "value": "src/claimflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/claimflow/compat.py", "compat.py"], "kind": "file", "value": "src/claimflow/compat.py"},...
[{"answers": ["src/claimflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01484
tileledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_fix grows linearly instead of doubling. It lives in src/tileledger/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tileledger layout. [tool_use kind=search nam...
67,032
3,072
100
115
false
synth:1172578311623600874@1.43875
synth
[{"aliases": ["src/tileledger/compat.py", "compat.py"], "kind": "file", "value": "src/tileledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tileledger/rules.py", "rules.py"], "kind": "file", "value": "src/tileledger/rules.py"}...
[{"answers": ["src/tileledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01484-twin
tileledger
ctx_compress-s1-01484
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_fix grows linearly instead of doubling. It lives in src/tileledger/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tileledger layout. [tool_use kind=search nam...
34,125
1,706
54
115
false
synth:1172578311623600874@1.43875
synth
[{"aliases": ["src/tileledger/compat.py", "compat.py"], "kind": "file", "value": "src/tileledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tileledger/rules.py", "rules.py"], "kind": "file", "value": "src/tileledger/rules.py"}...
[{"answers": ["src/tileledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01485
quotaflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_quota_batch double counts carried amounts when the period closes on a weekend. It lives in src/quotaflow/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the quotaflow l...
48,327
2,416
90
111
false
synth:6195242287621943666@1.25592
synth
[{"aliases": ["src/quotaflow/cli.py", "cli.py"], "kind": "file", "value": "src/quotaflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/quotaflow/store.py", "store.py"], "kind": "file", "value": "src/quotaflow/store.py"}, {"aliases": ["src/qu...
[{"answers": ["src/quotaflow/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rec...
ctx_compress-s1-01485-twin
quotaflow
ctx_compress-s1-01485
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_quota_batch double counts carried amounts when the period closes on a weekend. It lives in src/quotaflow/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the quotaflow l...
28,951
1,447
53
111
false
synth:6195242287621943666@1.25592
synth
[{"aliases": ["src/quotaflow/cli.py", "cli.py"], "kind": "file", "value": "src/quotaflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/quotaflow/store.py", "store.py"], "kind": "file", "value": "src/quotaflow/store.py"}, {"aliases": ["src/qu...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/quotaflow/core.py"], "...
ctx_compress-s1-01486
waypointflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_waypoint raises KeyError for a missing key instead of returning the default. It lives in src/waypointflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the wa...
28,097
1,404
52
65
false
synth:4863980421352075310@1
synth
[{"aliases": ["src/waypointflow/report.py", "report.py"], "kind": "file", "value": "src/waypointflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/waypointflow/schema.py", "schema.py"], "kind": "file", "value": "src/waypointflow/...
[{"answers": ["src/waypointflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01486-twin
waypointflow
ctx_compress-s1-01486
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_waypoint raises KeyError for a missing key instead of returning the default. It lives in src/waypointflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the wa...
19,960
1,000
38
65
false
synth:4863980421352075310@1
synth
[{"aliases": ["src/waypointflow/report.py", "report.py"], "kind": "file", "value": "src/waypointflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/waypointflow/schema.py", "schema.py"], "kind": "file", "value": "src/waypointflow/...
[{"answers": ["src/waypointflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01487
shipmentdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_pallet lets defaults overwrite explicit overrides. It lives in src/shipmentdesk/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the shipmentdesk layout. [tool_use kind=...
99,966
3,072
120
143
false
synth:8747406313298264725@1.95825
synth
[{"aliases": ["src/shipmentdesk/core.py", "core.py"], "kind": "file", "value": "src/shipmentdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/sku_0.csv", "sku_0.csv"], "kind": "file", "value": "data/sku_0.csv"}, {"aliases": ["data/reor...
[{"answers": ["src/shipmentdesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01487-twin
shipmentdesk
ctx_compress-s1-01487
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_pallet lets defaults overwrite explicit overrides. It lives in src/shipmentdesk/core.py; `python -m pytest tests/test_core.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the shipmentdesk layout. [tool_use kind=...
57,954
2,897
76
143
false
synth:8747406313298264725@1.95825
synth
[{"aliases": ["src/shipmentdesk/core.py", "core.py"], "kind": "file", "value": "src/shipmentdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/sku_0.csv", "sku_0.csv"], "kind": "file", "value": "data/sku_0.csv"}, {"aliases": ["data/reor...
[{"answers": ["src/shipmentdesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01488
calendarworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_quota grows linearly instead of doubling. It lives in src/calendarworks/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the calendarworks layout. [tool_use kind=...
54,108
2,705
95
110
false
synth:3220832359059459369@1.12315
synth
[{"aliases": ["src/calendarworks/compat.py", "compat.py"], "kind": "file", "value": "src/calendarworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/quota_0.csv", "quota_0.csv"], "kind": "file", "value": "data/quota_0.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01488-twin
calendarworks
ctx_compress-s1-01488
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_quota grows linearly instead of doubling. It lives in src/calendarworks/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the calendarworks layout. [tool_use kind=...
27,023
1,351
53
110
false
synth:3220832359059459369@1.12315
synth
[{"aliases": ["src/calendarworks/compat.py", "compat.py"], "kind": "file", "value": "src/calendarworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/quota_0.csv", "quota_0.csv"], "kind": "file", "value": "data/quota_0.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01489
tileforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_fix keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/tileforge/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tileforge lay...
64,515
3,072
101
108
false
synth:5007483223317198008@1.47305
synth
[{"aliases": ["src/tileforge/limits.py", "limits.py"], "kind": "file", "value": "src/tileforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/fix_5.csv", "fix_5.csv"], "kind": "file", "value": "data/fix_5.csv"}, {"aliases": ["src...
[{"answers": ["src/tileforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01489-twin
tileforge
ctx_compress-s1-01489
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_fix keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/tileforge/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the tileforge lay...
41,571
2,078
63
108
false
synth:5007483223317198008@1.47305
synth
[{"aliases": ["src/tileforge/limits.py", "limits.py"], "kind": "file", "value": "src/tileforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/fix_5.csv", "fix_5.csv"], "kind": "file", "value": "data/fix_5.csv"}, {"aliases": ["src...
[{"answers": ["src/tileforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-01490
warehousekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_stock averages one value too many once the window is full. It lives in src/warehousekit/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the warehousekit layout. [tool...
25,009
1,250
64
71
false
synth:6816938048088883136@1
synth
[{"aliases": ["src/warehousekit/cli.py", "cli.py"], "kind": "file", "value": "src/warehousekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/warehousekit/compat.py", "compat.py"], "kind": "file", "value": "src/warehousekit/compat.py"}, {"ali...
[{"answers": ["src/warehousekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01490-twin
warehousekit
ctx_compress-s1-01490
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_stock averages one value too many once the window is full. It lives in src/warehousekit/cli.py; `python -m pytest tests/test_cli.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the warehousekit layout. [tool...
14,915
1,000
38
71
false
synth:6816938048088883136@1
synth
[{"aliases": ["src/warehousekit/cli.py", "cli.py"], "kind": "file", "value": "src/warehousekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/warehousekit/compat.py", "compat.py"], "kind": "file", "value": "src/warehousekit/compat.py"}, {"ali...
[{"answers": ["src/warehousekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01491
stockdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_lot ignores the cap, so late attempts wait for minutes. It lives in src/stockdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockdesk layout. [tool_use ki...
44,007
2,200
60
71
false
synth:4418536066495435876@1
synth
[{"aliases": ["src/stockdesk/format.py", "format.py"], "kind": "file", "value": "src/stockdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/stock_0.csv", "stock_0.csv"], "kind": "file", "value": "data/stock_0.csv"}, {"aliases":...
[{"answers": ["src/stockdesk/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01491-twin
stockdesk
ctx_compress-s1-01491
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_lot ignores the cap, so late attempts wait for minutes. It lives in src/stockdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the stockdesk layout. [tool_use ki...
19,556
1,000
35
71
false
synth:4418536066495435876@1
synth
[{"aliases": ["src/stockdesk/format.py", "format.py"], "kind": "file", "value": "src/stockdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/stock_0.csv", "stock_0.csv"], "kind": "file", "value": "data/stock_0.csv"}, {"aliases":...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/stockdesk/format.py"...
ctx_compress-s1-01492
policyflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_principal produces doubled dashes for consecutive separators. It lives in src/policyflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the policyflow layout. [tool_use...
56,500
2,825
82
91
false
synth:3411264351642637643@1
synth
[{"aliases": ["src/policyflow/io.py", "io.py"], "kind": "file", "value": "src/policyflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/policyflow/schema.py", "schema.py"], "kind": "file", "value": "src/policyflow/schema.py"}, {"aliases": ["data/...
[{"answers": ["src/policyflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-01492-twin
policyflow
ctx_compress-s1-01492
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_principal produces doubled dashes for consecutive separators. It lives in src/policyflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the policyflow layout. [tool_use...
28,152
1,407
46
91
false
synth:3411264351642637643@1
synth
[{"aliases": ["src/policyflow/io.py", "io.py"], "kind": "file", "value": "src/policyflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/policyflow/schema.py", "schema.py"], "kind": "file", "value": "src/policyflow/schema.py"}, {"aliases": ["data/...
[{"answers": ["src/policyflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-01493
spanforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_counter_batch raises ValueError when a value itself contains an equals sign. It lives in src/spanforge/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the spanforg...
56,967
2,848
103
110
false
synth:6737041761482837229@1.30828
synth
[{"aliases": ["src/spanforge/format.py", "format.py"], "kind": "file", "value": "src/spanforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/spanforge/io.py", "io.py"], "kind": "file", "value": "src/spanforge/io.py"}, {"aliases":...
[{"answers": ["src/spanforge/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01493-twin
spanforge
ctx_compress-s1-01493
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_counter_batch raises ValueError when a value itself contains an equals sign. It lives in src/spanforge/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the spanforg...
29,941
1,497
57
110
false
synth:6737041761482837229@1.30828
synth
[{"aliases": ["src/spanforge/format.py", "format.py"], "kind": "file", "value": "src/spanforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/spanforge/io.py", "io.py"], "kind": "file", "value": "src/spanforge/io.py"}, {"aliases":...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["audit.py", "src/spanforge/audit.py"],...
ctx_compress-s1-01494
cueforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_cue_window ignores the cap, so late attempts wait for minutes. It lives in src/cueforge/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cueforge layout. [...
69,655
3,072
80
95
false
synth:5717605979840387987@1.28602
synth
[{"aliases": ["src/cueforge/limits.py", "limits.py"], "kind": "file", "value": "src/cueforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/cueforge/report.py", "report.py"], "kind": "file", "value": "src/cueforge/report.py"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/cueforge/tables.py", "tables.py"]...
ctx_compress-s1-01494-twin
cueforge
ctx_compress-s1-01494
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_cue_window ignores the cap, so late attempts wait for minutes. It lives in src/cueforge/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cueforge layout. [...
44,665
2,233
52
95
false
synth:5717605979840387987@1.28602
synth
[{"aliases": ["src/cueforge/limits.py", "limits.py"], "kind": "file", "value": "src/cueforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/cueforge/report.py", "report.py"], "kind": "file", "value": "src/cueforge/report.py"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/stem_6.csv", "stem_6.csv"], "kin...
ctx_compress-s1-01495
revisionworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index grows linearly instead of doubling. It lives in src/revisionworks/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the revisionworks layout. [tool_use kin...
52,216
2,610
95
114
false
synth:3368721486728670240@1.1438
synth
[{"aliases": ["src/revisionworks/limits.py", "limits.py"], "kind": "file", "value": "src/revisionworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/revisionworks/format.py", "format.py"], "kind": "file", "value": "src/revisionwo...
[{"answers": ["src/revisionworks/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent'...
ctx_compress-s1-01495-twin
revisionworks
ctx_compress-s1-01495
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index grows linearly instead of doubling. It lives in src/revisionworks/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the revisionworks layout. [tool_use kin...
33,982
1,699
59
114
false
synth:3368721486728670240@1.1438
synth
[{"aliases": ["src/revisionworks/limits.py", "limits.py"], "kind": "file", "value": "src/revisionworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/revisionworks/format.py", "format.py"], "kind": "file", "value": "src/revisionwo...
[{"answers": ["src/revisionworks/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent'...
ctx_compress-s1-01496
clausekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_index keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/clausekit/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the clausekit layou...
68,614
3,072
92
109
false
synth:4485811543924157236@1.73425
synth
[{"aliases": ["src/clausekit/audit.py", "audit.py"], "kind": "file", "value": "src/clausekit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/section_5.csv", "section_5.csv"], "kind": "file", "value": "data/section_5.csv"}, {"aliases":...
[{"answers": ["src/clausekit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01496-twin
clausekit
ctx_compress-s1-01496
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_index keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/clausekit/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the clausekit layou...
34,809
1,740
58
109
false
synth:4485811543924157236@1.73425
synth
[{"aliases": ["src/clausekit/audit.py", "audit.py"], "kind": "file", "value": "src/clausekit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/section_5.csv", "section_5.csv"], "kind": "file", "value": "data/section_5.csv"}, {"aliases":...
[{"answers": ["src/clausekit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01497
bearingledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_zone raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/bearingledger/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
67,585
3,072
91
106
false
synth:4151095231118804105@1.2207
synth
[{"aliases": ["src/bearingledger/limits.py", "limits.py"], "kind": "file", "value": "src/bearingledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/bearingledger/compat.py", "compat.py"], "kind": "file", "value": "src/bearingled...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["schema.py", "src/bearingledger/schema...
ctx_compress-s1-01497-twin
bearingledger
ctx_compress-s1-01497
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_zone raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/bearingledger/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
44,255
2,212
59
106
false
synth:4151095231118804105@1.2207
synth
[{"aliases": ["src/bearingledger/limits.py", "limits.py"], "kind": "file", "value": "src/bearingledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/bearingledger/compat.py", "compat.py"], "kind": "file", "value": "src/bearingled...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/leg_6.csv", "leg_6.csv"], "kind"...
ctx_compress-s1-01498
slotledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_calendar double counts carried amounts when the period closes on a weekend. It lives in src/slotledger/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the slotledg...
110,965
3,072
108
127
false
synth:1225755690486065452@2.03405
synth
[{"aliases": ["src/slotledger/rules.py", "rules.py"], "kind": "file", "value": "src/slotledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/slot_7.csv", "slot_7.csv"], "kind": "file", "value": "data/slot_7.csv"}, {"aliases": ["data...
[{"answers": ["src/slotledger/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01498-twin
slotledger
ctx_compress-s1-01498
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_calendar double counts carried amounts when the period closes on a weekend. It lives in src/slotledger/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the slotledg...
74,172
3,072
76
127
false
synth:1225755690486065452@2.03405
synth
[{"aliases": ["src/slotledger/rules.py", "rules.py"], "kind": "file", "value": "src/slotledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/slot_7.csv", "slot_7.csv"], "kind": "file", "value": "data/slot_7.csv"}, {"aliases": ["data...
[{"answers": ["src/slotledger/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01499
indexledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_batch raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/indexledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping...
69,915
3,072
90
113
false
synth:7011901973515463728@1.63258
synth
[{"aliases": ["src/indexledger/report.py", "report.py"], "kind": "file", "value": "src/indexledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/indexledger/tables.py", "tables.py"], "kind": "file", "value": "src/indexledger/tabl...
[{"answers": ["src/indexledger/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01499-twin
indexledger
ctx_compress-s1-01499
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_batch raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/indexledger/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping...
50,797
2,539
66
113
false
synth:7011901973515463728@1.63258
synth
[{"aliases": ["src/indexledger/report.py", "report.py"], "kind": "file", "value": "src/indexledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/indexledger/tables.py", "tables.py"], "kind": "file", "value": "src/indexledger/tabl...
[{"answers": ["src/indexledger/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01500
claimdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_principal keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/claimdesk/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the claimdesk...
31,348
1,567
52
59
false
synth:4981418727050310921@1
synth
[{"aliases": ["src/claimdesk/limits.py", "limits.py"], "kind": "file", "value": "src/claimdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/claimdesk/report.py", "report.py"], "kind": "file", "value": "src/claimdesk/report.py"},...
[{"answers": ["src/claimdesk/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01500-twin
claimdesk
ctx_compress-s1-01500
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_principal keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/claimdesk/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the claimdesk...
18,102
1,000
30
59
false
synth:4981418727050310921@1
synth
[{"aliases": ["src/claimdesk/limits.py", "limits.py"], "kind": "file", "value": "src/claimdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/claimdesk/report.py", "report.py"], "kind": "file", "value": "src/claimdesk/report.py"},...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["compat.py", "src/claimdesk/compat.py"...
ctx_compress-s1-01501
cadenceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_quota ignores the cap, so late attempts wait for minutes. It lives in src/cadenceledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cadenceledger layout...
42,541
2,127
78
95
false
synth:7937585192371264698@1.21307
synth
[{"aliases": ["src/cadenceledger/format.py", "format.py"], "kind": "file", "value": "src/cadenceledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/cadenceledger/report.py", "report.py"], "kind": "file", "value": "src/cadenceled...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["compat.py", "src/cadenceledger/compat...
ctx_compress-s1-01501-twin
cadenceledger
ctx_compress-s1-01501
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_quota ignores the cap, so late attempts wait for minutes. It lives in src/cadenceledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the cadenceledger layout...
24,389
1,219
45
95
false
synth:7937585192371264698@1.21307
synth
[{"aliases": ["src/cadenceledger/format.py", "format.py"], "kind": "file", "value": "src/cadenceledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/cadenceledger/report.py", "report.py"], "kind": "file", "value": "src/cadenceled...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["booking_2.csv", "data/booking_2.csv"]...
ctx_compress-s1-01502
clauseledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index_batch double counts carried amounts when the period closes on a weekend. It lives in src/clauseledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the...
42,951
2,147
58
71
false
synth:1842939606403589725@1
synth
[{"aliases": ["src/clauseledger/format.py", "format.py"], "kind": "file", "value": "src/clauseledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/heading_3.csv", "heading_3.csv"], "kind": "file", "value": "data/heading_3.csv"},...
[{"answers": ["src/clauseledger/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01502-twin
clauseledger
ctx_compress-s1-01502
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index_batch double counts carried amounts when the period closes on a weekend. It lives in src/clauseledger/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the...
20,516
1,025
32
71
false
synth:1842939606403589725@1
synth
[{"aliases": ["src/clauseledger/format.py", "format.py"], "kind": "file", "value": "src/clauseledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/heading_3.csv", "heading_3.csv"], "kind": "file", "value": "data/heading_3.csv"},...
[{"answers": ["src/clauseledger/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01503
captionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_slug_safe returns the upper bound for every value instead of clamping into the range. It lives in src/captionkit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mappin...
89,806
3,072
114
135
false
synth:6459304022387181436@1.596
synth
[{"aliases": ["src/captionkit/limits.py", "limits.py"], "kind": "file", "value": "src/captionkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/captionkit/io.py", "io.py"], "kind": "file", "value": "src/captionkit/io.py"}, {"alias...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_rules.py", "tests/test_rules.py"...
ctx_compress-s1-01503-twin
captionkit
ctx_compress-s1-01503
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_slug_safe returns the upper bound for every value instead of clamping into the range. It lives in src/captionkit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mappin...
57,359
2,867
74
135
false
synth:6459304022387181436@1.596
synth
[{"aliases": ["src/captionkit/limits.py", "limits.py"], "kind": "file", "value": "src/captionkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/captionkit/io.py", "io.py"], "kind": "file", "value": "src/captionkit/io.py"}, {"alias...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["report.py", "src/captionkit/report.py...
ctx_compress-s1-01504
ledgerflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_credit double counts carried amounts when the period closes on a weekend. It lives in src/ledgerflow/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ledgerflow...
79,672
3,072
110
133
false
synth:4985077002256447858@2.05945
synth
[{"aliases": ["src/ledgerflow/compat.py", "compat.py"], "kind": "file", "value": "src/ledgerflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/ledgerflow/format.py", "format.py"], "kind": "file", "value": "src/ledgerflow/format.p...
[{"answers": ["src/ledgerflow/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01504-twin
ledgerflow
ctx_compress-s1-01504
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_credit double counts carried amounts when the period closes on a weekend. It lives in src/ledgerflow/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ledgerflow...
41,025
2,051
65
133
false
synth:4985077002256447858@2.05945
synth
[{"aliases": ["src/ledgerflow/compat.py", "compat.py"], "kind": "file", "value": "src/ledgerflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/ledgerflow/format.py", "format.py"], "kind": "file", "value": "src/ledgerflow/format.p...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_format.py", "tests/test_format.p...
ctx_compress-s1-01505
ledgerdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_payout_once rejects a range where lo equals hi, which is a valid single point range. It lives in src/ledgerdesk/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the...
43,938
2,196
80
97
false
synth:2171807256343189262@1.22085
synth
[{"aliases": ["src/ledgerdesk/store.py", "store.py"], "kind": "file", "value": "src/ledgerdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/ledgerdesk/audit.py", "audit.py"], "kind": "file", "value": "src/ledgerdesk/audit.py"}, {"al...
[{"answers": ["src/ledgerdesk/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01505-twin
ledgerdesk
ctx_compress-s1-01505
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_payout_once rejects a range where lo equals hi, which is a valid single point range. It lives in src/ledgerdesk/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the...
29,176
1,458
57
97
false
synth:2171807256343189262@1.22085
synth
[{"aliases": ["src/ledgerdesk/store.py", "store.py"], "kind": "file", "value": "src/ledgerdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/ledgerdesk/audit.py", "audit.py"], "kind": "file", "value": "src/ledgerdesk/audit.py"}, {"al...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01506
waypointflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] bound_bearing grows linearly instead of doubling. It lives in src/waypointflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the waypointflow layout. [tool_use kind=search na...
90,287
3,072
112
121
false
synth:3314557450850600083@1.69233
synth
[{"aliases": ["src/waypointflow/io.py", "io.py"], "kind": "file", "value": "src/waypointflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/corridor_8.csv", "corridor_8.csv"], "kind": "file", "value": "data/corridor_8.csv"}, {"aliases": ["src/wa...
[{"answers": ["src/waypointflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-01506-twin
waypointflow
ctx_compress-s1-01506
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] bound_bearing grows linearly instead of doubling. It lives in src/waypointflow/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the waypointflow layout. [tool_use kind=search na...
48,738
2,436
58
121
false
synth:3314557450850600083@1.69233
synth
[{"aliases": ["src/waypointflow/io.py", "io.py"], "kind": "file", "value": "src/waypointflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/corridor_8.csv", "corridor_8.csv"], "kind": "file", "value": "data/corridor_8.csv"}, {"aliases": ["src/wa...
[{"answers": ["src/waypointflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most re...