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-s1-00955-twin
nonceflow
ctx_compress-s1-00955
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_policy_window lets defaults overwrite explicit overrides. It lives in src/nonceflow/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 nonceflow layout. [tool_use...
33,358
1,667
55
119
false
synth:8067359535806867614@1.4475
synth
[{"aliases": ["src/nonceflow/rules.py", "rules.py"], "kind": "file", "value": "src/nonceflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/nonceflow/compat.py", "compat.py"], "kind": "file", "value": "src/nonceflow/compat.py"}, {"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": ["compat.py", "src/nonceflow/compat.py"...
ctx_compress-s1-00956
zonedesk
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_tile_once drops override values of 0, False and empty string. It lives in src/zonedesk/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 zonedesk layout. [tool_use k...
26,318
1,315
53
60
false
synth:214338799586223457@1
synth
[{"aliases": ["src/zonedesk/io.py", "io.py"], "kind": "file", "value": "src/zonedesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/beacon_1.csv", "beacon_1.csv"], "kind": "file", "value": "data/beacon_1.csv"}, {"aliases": ["src/zonedesk/core.py...
[{"answers": ["src/zonedesk/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 recent...
ctx_compress-s1-00956-twin
zonedesk
ctx_compress-s1-00956
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_tile_once drops override values of 0, False and empty string. It lives in src/zonedesk/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 zonedesk layout. [tool_use k...
14,441
1,000
33
60
false
synth:214338799586223457@1
synth
[{"aliases": ["src/zonedesk/io.py", "io.py"], "kind": "file", "value": "src/zonedesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/beacon_1.csv", "beacon_1.csv"], "kind": "file", "value": "data/beacon_1.csv"}, {"aliases": ["src/zonedesk/core.py...
[{"answers": ["src/zonedesk/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-s1-00957
indexforge
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_clause raises DraftBoundError for bytes input in strict mode instead of decoding it. It lives in src/indexforge/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...
61,487
3,072
96
109
false
synth:8793648620721227632@1.09425
synth
[{"aliases": ["src/indexforge/store.py", "store.py"], "kind": "file", "value": "src/indexforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/indexforge/io.py", "io.py"], "kind": "file", "value": "src/indexforge/io.py"}, {"aliases": [...
[{"answers": ["src/indexforge/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-s1-00957-twin
indexforge
ctx_compress-s1-00957
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_clause raises DraftBoundError for bytes input in strict mode instead of decoding it. It lives in src/indexforge/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...
32,189
1,609
56
109
false
synth:8793648620721227632@1.09425
synth
[{"aliases": ["src/indexforge/store.py", "store.py"], "kind": "file", "value": "src/indexforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/indexforge/io.py", "io.py"], "kind": "file", "value": "src/indexforge/io.py"}, {"aliases": [...
[{"answers": ["src/indexforge/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 mos...
ctx_compress-s1-00958
slugdesk
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_snippet lets defaults overwrite explicit overrides. It lives in src/slugdesk/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 slugdesk layout. [tool_use kind...
99,238
3,072
134
149
false
synth:6814060659867281489@2.0219
synth
[{"aliases": ["src/slugdesk/rules.py", "rules.py"], "kind": "file", "value": "src/slugdesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/stem_3.csv", "stem_3.csv"], "kind": "file", "value": "data/stem_3.csv"}, {"aliases": ["src/slug...
[{"answers": ["src/slugdesk/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 most ...
ctx_compress-s1-00958-twin
slugdesk
ctx_compress-s1-00958
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_snippet lets defaults overwrite explicit overrides. It lives in src/slugdesk/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 slugdesk layout. [tool_use kind...
58,565
2,928
82
149
false
synth:6814060659867281489@2.0219
synth
[{"aliases": ["src/slugdesk/rules.py", "rules.py"], "kind": "file", "value": "src/slugdesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/stem_3.csv", "stem_3.csv"], "kind": "file", "value": "data/stem_3.csv"}, {"aliases": ["src/slug...
[{"answers": ["src/slugdesk/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 most ...
ctx_compress-s1-00959
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] transfer_lot lets defaults overwrite explicit overrides. It lives in src/warehousekit/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 warehousekit layout. [tool_use...
86,595
3,072
101
120
false
synth:8051292366574103305@1.66817
synth
[{"aliases": ["src/warehousekit/report.py", "report.py"], "kind": "file", "value": "src/warehousekit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/warehousekit/cli.py", "cli.py"], "kind": "file", "value": "src/warehousekit/cli.py...
[{"answers": ["src/warehousekit/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-s1-00959-twin
warehousekit
ctx_compress-s1-00959
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_lot lets defaults overwrite explicit overrides. It lives in src/warehousekit/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 warehousekit layout. [tool_use...
50,749
2,537
69
120
false
synth:8051292366574103305@1.66817
synth
[{"aliases": ["src/warehousekit/report.py", "report.py"], "kind": "file", "value": "src/warehousekit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/warehousekit/cli.py", "cli.py"], "kind": "file", "value": "src/warehousekit/cli.py...
[{"answers": ["src/warehousekit/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-s1-00960
grantforge
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 skips the first page; page 1 returns what should be page 2. It lives in src/grantforge/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 grantforge layout....
72,892
3,072
106
121
false
synth:7063881381638838539@2.30748
synth
[{"aliases": ["src/grantforge/store.py", "store.py"], "kind": "file", "value": "src/grantforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/claim_1.csv", "claim_1.csv"], "kind": "file", "value": "data/claim_1.csv"}, {"aliases": ["s...
[{"answers": ["src/grantforge/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-s1-00960-twin
grantforge
ctx_compress-s1-00960
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 skips the first page; page 1 returns what should be page 2. It lives in src/grantforge/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 grantforge layout....
38,296
1,914
59
121
false
synth:7063881381638838539@2.30748
synth
[{"aliases": ["src/grantforge/store.py", "store.py"], "kind": "file", "value": "src/grantforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/claim_1.csv", "claim_1.csv"], "kind": "file", "value": "data/claim_1.csv"}, {"aliases": ["s...
[{"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_limits.py", "tests/test_limits.p...
ctx_compress-s1-00961
lotkit
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 leaves a trailing dash when the title ends in punctuation. It lives in src/lotkit/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 lotkit layout. [tool_...
57,166
2,858
95
104
false
synth:920383812367982373@1.04555
synth
[{"aliases": ["src/lotkit/limits.py", "limits.py"], "kind": "file", "value": "src/lotkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/warehouse_2.csv", "warehouse_2.csv"], "kind": "file", "value": "data/warehouse_2.csv"}, {"ali...
[{"answers": ["src/lotkit/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 most ...
ctx_compress-s1-00961-twin
lotkit
ctx_compress-s1-00961
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 leaves a trailing dash when the title ends in punctuation. It lives in src/lotkit/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 lotkit layout. [tool_...
33,486
1,674
59
104
false
synth:920383812367982373@1.04555
synth
[{"aliases": ["src/lotkit/limits.py", "limits.py"], "kind": "file", "value": "src/lotkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/warehouse_2.csv", "warehouse_2.csv"], "kind": "file", "value": "data/warehouse_2.csv"}, {"ali...
[{"answers": ["src/lotkit/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 most ...
ctx_compress-s1-00962
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] pick_lot raises ValueError when a value itself contains an equals sign. It lives in src/warehousekit/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 warehousekit layout...
49,057
2,452
90
113
false
synth:1435004997744542113@1.04792
synth
[{"aliases": ["src/warehousekit/core.py", "core.py"], "kind": "file", "value": "src/warehousekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/warehousekit/store.py", "store.py"], "kind": "file", "value": "src/warehousekit/store.py"}, {"...
[{"answers": ["src/warehousekit/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-s1-00962-twin
warehousekit
ctx_compress-s1-00962
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_lot raises ValueError when a value itself contains an equals sign. It lives in src/warehousekit/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 warehousekit layout...
30,009
1,500
53
113
false
synth:1435004997744542113@1.04792
synth
[{"aliases": ["src/warehousekit/core.py", "core.py"], "kind": "file", "value": "src/warehousekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/warehousekit/store.py", "store.py"], "kind": "file", "value": "src/warehousekit/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": ["core.py", "src/warehousekit/core.py"]...
ctx_compress-s1-00963
policydesk
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_scope skips the first page; page 1 returns what should be page 2. It lives in src/policydesk/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 policydesk layout....
72,063
3,072
92
105
false
synth:6090786770289336282@1.20245
synth
[{"aliases": ["src/policydesk/audit.py", "audit.py"], "kind": "file", "value": "src/policydesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policydesk/tables.py", "tables.py"], "kind": "file", "value": "src/policydesk/tables.py"}, {...
[{"answers": ["src/policydesk/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-s1-00963-twin
policydesk
ctx_compress-s1-00963
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_scope skips the first page; page 1 returns what should be page 2. It lives in src/policydesk/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 policydesk layout....
35,131
1,756
48
105
false
synth:6090786770289336282@1.20245
synth
[{"aliases": ["src/policydesk/audit.py", "audit.py"], "kind": "file", "value": "src/policydesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policydesk/tables.py", "tables.py"], "kind": "file", "value": "src/policydesk/tables.py"}, {...
[{"answers": ["src/policydesk/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": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-00964
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 returns one item too few on every page. It lives in src/clauseledger/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 clauseledger layout. [tool_use k...
33,032
1,651
72
87
false
synth:5361854185950856635@1
synth
[{"aliases": ["src/clauseledger/compat.py", "compat.py"], "kind": "file", "value": "src/clauseledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/clauseledger/limits.py", "limits.py"], "kind": "file", "value": "src/clauseledger/...
[{"answers": ["src/clauseledger/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...
ctx_compress-s1-00964-twin
clauseledger
ctx_compress-s1-00964
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 returns one item too few on every page. It lives in src/clauseledger/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 clauseledger layout. [tool_use k...
19,701
1,000
52
87
false
synth:5361854185950856635@1
synth
[{"aliases": ["src/clauseledger/compat.py", "compat.py"], "kind": "file", "value": "src/clauseledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/clauseledger/limits.py", "limits.py"], "kind": "file", "value": "src/clauseledger/...
[{"answers": ["src/clauseledger/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...
ctx_compress-s1-00965
marginforge
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_revision keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/marginforge/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 margin...
59,392
2,969
100
107
false
synth:8004780750817935496@1.457
synth
[{"aliases": ["src/marginforge/report.py", "report.py"], "kind": "file", "value": "src/marginforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/marginforge/schema.py", "schema.py"], "kind": "file", "value": "src/marginforge/sche...
[{"answers": ["src/marginforge/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-s1-00965-twin
marginforge
ctx_compress-s1-00965
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_revision keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/marginforge/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 margin...
32,366
1,618
49
107
false
synth:8004780750817935496@1.457
synth
[{"aliases": ["src/marginforge/report.py", "report.py"], "kind": "file", "value": "src/marginforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/marginforge/schema.py", "schema.py"], "kind": "file", "value": "src/marginforge/sche...
[{"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/marginforge/tables.py", "tables.p...
ctx_compress-s1-00966
slugforge
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_caption raises KeyError for a missing key instead of returning the default. It lives in src/slugforge/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 slug...
60,382
3,019
90
105
false
synth:758049979782794937@1.60023
synth
[{"aliases": ["src/slugforge/format.py", "format.py"], "kind": "file", "value": "src/slugforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slugforge/io.py", "io.py"], "kind": "file", "value": "src/slugforge/io.py"}, {"aliases":...
[{"answers": ["src/slugforge/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 mo...
ctx_compress-s1-00966-twin
slugforge
ctx_compress-s1-00966
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_caption raises KeyError for a missing key instead of returning the default. It lives in src/slugforge/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 slug...
29,754
1,487
56
105
false
synth:758049979782794937@1.60023
synth
[{"aliases": ["src/slugforge/format.py", "format.py"], "kind": "file", "value": "src/slugforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slugforge/io.py", "io.py"], "kind": "file", "value": "src/slugforge/io.py"}, {"aliases":...
[{"answers": ["src/slugforge/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-s1-00967
calendarflow
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_slot_safe rejects a range where lo equals hi, which is a valid single point range. It lives in src/calendarflow/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 ...
69,358
3,072
84
107
false
synth:5589936017715791920@1.766
synth
[{"aliases": ["src/calendarflow/report.py", "report.py"], "kind": "file", "value": "src/calendarflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/calendarflow/audit.py", "audit.py"], "kind": "file", "value": "src/calendarflow/au...
[{"answers": ["src/calendarflow/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-s1-00967-twin
calendarflow
ctx_compress-s1-00967
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_slot_safe rejects a range where lo equals hi, which is a valid single point range. It lives in src/calendarflow/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 ...
42,641
2,132
60
107
false
synth:5589936017715791920@1.766
synth
[{"aliases": ["src/calendarflow/report.py", "report.py"], "kind": "file", "value": "src/calendarflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/calendarflow/audit.py", "audit.py"], "kind": "file", "value": "src/calendarflow/au...
[{"answers": ["src/calendarflow/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-s1-00968
chunkledger
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] truncate_cue returns the upper bound for every value instead of clamping into the range. It lives in src/chunkledger/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...
46,091
2,304
70
83
false
synth:6791068002344136392@1
synth
[{"aliases": ["src/chunkledger/compat.py", "compat.py"], "kind": "file", "value": "src/chunkledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/chunkledger/report.py", "report.py"], "kind": "file", "value": "src/chunkledger/repo...
[{"answers": ["src/chunkledger/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 ...
ctx_compress-s1-00968-twin
chunkledger
ctx_compress-s1-00968
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] truncate_cue returns the upper bound for every value instead of clamping into the range. It lives in src/chunkledger/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...
22,030
1,101
42
83
false
synth:6791068002344136392@1
synth
[{"aliases": ["src/chunkledger/compat.py", "compat.py"], "kind": "file", "value": "src/chunkledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/chunkledger/report.py", "report.py"], "kind": "file", "value": "src/chunkledger/repo...
[{"answers": ["src/chunkledger/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 ...
ctx_compress-s1-00969
skudesk
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_sku raises ValueError when a value itself contains an equals sign. It lives in src/skudesk/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 skudesk layout. ...
38,798
1,939
74
83
false
synth:3307610455371300726@1
synth
[{"aliases": ["src/skudesk/format.py", "format.py"], "kind": "file", "value": "src/skudesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/skudesk/compat.py", "compat.py"], "kind": "file", "value": "src/skudesk/compat.py"}, {"alias...
[{"answers": ["src/skudesk/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 most...
ctx_compress-s1-00969-twin
skudesk
ctx_compress-s1-00969
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_sku raises ValueError when a value itself contains an equals sign. It lives in src/skudesk/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 skudesk layout. ...
20,631
1,031
39
83
false
synth:3307610455371300726@1
synth
[{"aliases": ["src/skudesk/format.py", "format.py"], "kind": "file", "value": "src/skudesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/skudesk/compat.py", "compat.py"], "kind": "file", "value": "src/skudesk/compat.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": ["bin_3.csv", "data/bin_3.csv"], "kind"...
ctx_compress-s1-00970
reorderledger
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_shipment rejects a range where lo equals hi, which is a valid single point range. It lives in src/reorderledger/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 reorder...
26,986
1,349
64
79
false
synth:589200883522864163@1
synth
[{"aliases": ["src/reorderledger/io.py", "io.py"], "kind": "file", "value": "src/reorderledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/reorderledger/compat.py", "compat.py"], "kind": "file", "value": "src/reorderledger/compat.py"}, {"alias...
[{"answers": ["src/reorderledger/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 r...
ctx_compress-s1-00970-twin
reorderledger
ctx_compress-s1-00970
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_shipment rejects a range where lo equals hi, which is a valid single point range. It lives in src/reorderledger/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 reorder...
11,474
1,000
37
79
false
synth:589200883522864163@1
synth
[{"aliases": ["src/reorderledger/io.py", "io.py"], "kind": "file", "value": "src/reorderledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/reorderledger/compat.py", "compat.py"], "kind": "file", "value": "src/reorderledger/compat.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": ["data/sku_3.csv", "sku_3.csv"], "kind"...
ctx_compress-s1-00971
margindesk
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_revision returns the upper bound for every value instead of clamping into the range. It lives in src/margindesk/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...
47,001
2,350
66
79
false
synth:802463003617790780@1
synth
[{"aliases": ["src/margindesk/audit.py", "audit.py"], "kind": "file", "value": "src/margindesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/heading_5.csv", "heading_5.csv"], "kind": "file", "value": "data/heading_5.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/margindesk/tables.py", "tables.py...
ctx_compress-s1-00971-twin
margindesk
ctx_compress-s1-00971
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_revision returns the upper bound for every value instead of clamping into the range. It lives in src/margindesk/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...
23,338
1,166
36
79
false
synth:802463003617790780@1
synth
[{"aliases": ["src/margindesk/audit.py", "audit.py"], "kind": "file", "value": "src/margindesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/heading_5.csv", "heading_5.csv"], "kind": "file", "value": "data/heading_5.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": ["cli.py", "src/margindesk/cli.py"], "k...
ctx_compress-s1-00972
bucketledger
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_trace drops override values of 0, False and empty string. It lives in src/bucketledger/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 bucketledger layout. [tool_use...
89,199
3,072
98
121
false
synth:2866664183289353752@1.8319
synth
[{"aliases": ["src/bucketledger/cli.py", "cli.py"], "kind": "file", "value": "src/bucketledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/bucketledger/schema.py", "schema.py"], "kind": "file", "value": "src/bucketledger/schema.py"}, {"ali...
[{"answers": ["src/bucketledger/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-s1-00972-twin
bucketledger
ctx_compress-s1-00972
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_trace drops override values of 0, False and empty string. It lives in src/bucketledger/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 bucketledger layout. [tool_use...
49,645
2,482
60
121
false
synth:2866664183289353752@1.8319
synth
[{"aliases": ["src/bucketledger/cli.py", "cli.py"], "kind": "file", "value": "src/bucketledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/bucketledger/schema.py", "schema.py"], "kind": "file", "value": "src/bucketledger/schema.py"}, {"ali...
[{"answers": ["src/bucketledger/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-s1-00973
lexemeforge
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_once produces doubled dashes for consecutive separators. It lives in src/lexemeforge/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 lexemeforge layou...
98,896
3,072
112
131
false
synth:2206446517391195991@1.74085
synth
[{"aliases": ["src/lexemeforge/format.py", "format.py"], "kind": "file", "value": "src/lexemeforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/chunk_3.csv", "chunk_3.csv"], "kind": "file", "value": "data/chunk_3.csv"}, {"alias...
[{"answers": ["src/lexemeforge/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-s1-00973-twin
lexemeforge
ctx_compress-s1-00973
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_once produces doubled dashes for consecutive separators. It lives in src/lexemeforge/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 lexemeforge layou...
59,765
2,988
72
131
false
synth:2206446517391195991@1.74085
synth
[{"aliases": ["src/lexemeforge/format.py", "format.py"], "kind": "file", "value": "src/lexemeforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/chunk_3.csv", "chunk_3.csv"], "kind": "file", "value": "data/chunk_3.csv"}, {"alias...
[{"answers": ["src/lexemeforge/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-s1-00974
snippetflow
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_glyph averages one value too many once the window is full. It lives in src/snippetflow/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 snippetflow layout. [too...
94,689
3,072
99
112
false
synth:5867565799559880914@1.69477
synth
[{"aliases": ["src/snippetflow/core.py", "core.py"], "kind": "file", "value": "src/snippetflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/slug_1.csv", "slug_1.csv"], "kind": "file", "value": "data/slug_1.csv"}, {"aliases": ["data/ste...
[{"answers": ["src/snippetflow/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...
ctx_compress-s1-00974-twin
snippetflow
ctx_compress-s1-00974
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_glyph averages one value too many once the window is full. It lives in src/snippetflow/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 snippetflow layout. [too...
50,246
2,512
57
112
false
synth:5867565799559880914@1.69477
synth
[{"aliases": ["src/snippetflow/core.py", "core.py"], "kind": "file", "value": "src/snippetflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/slug_1.csv", "slug_1.csv"], "kind": "file", "value": "data/slug_1.csv"}, {"aliases": ["data/ste...
[{"answers": ["src/snippetflow/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-s1-00975
palletflow
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_bin_strict rejects a range where lo equals hi, which is a valid single point range. It lives in src/palletflow/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...
47,060
2,353
86
109
false
synth:5438634403881077024@1
synth
[{"aliases": ["src/palletflow/store.py", "store.py"], "kind": "file", "value": "src/palletflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/sku_0.csv", "sku_0.csv"], "kind": "file", "value": "data/sku_0.csv"}, {"aliases": ["src/pal...
[{"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/stock_5.csv", "stock_5.csv"], "k...
ctx_compress-s1-00975-twin
palletflow
ctx_compress-s1-00975
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_bin_strict rejects a range where lo equals hi, which is a valid single point range. It lives in src/palletflow/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...
39,148
1,957
62
109
false
synth:5438634403881077024@1
synth
[{"aliases": ["src/palletflow/store.py", "store.py"], "kind": "file", "value": "src/palletflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/sku_0.csv", "sku_0.csv"], "kind": "file", "value": "data/sku_0.csv"}, {"aliases": ["src/pal...
[{"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/reorder_2.csv", "reorder_2.csv"]...
ctx_compress-s1-00976
footnoteworks
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_clause_safe drops override values of 0, False and empty string. It lives in src/footnoteworks/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 footnoteworks l...
51,223
2,561
82
99
false
synth:8665533671581681023@1
synth
[{"aliases": ["src/footnoteworks/limits.py", "limits.py"], "kind": "file", "value": "src/footnoteworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/footnoteworks/report.py", "report.py"], "kind": "file", "value": "src/footnotewo...
[{"answers": ["src/footnoteworks/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-s1-00976-twin
footnoteworks
ctx_compress-s1-00976
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_clause_safe drops override values of 0, False and empty string. It lives in src/footnoteworks/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 footnoteworks l...
33,886
1,694
58
99
false
synth:8665533671581681023@1
synth
[{"aliases": ["src/footnoteworks/limits.py", "limits.py"], "kind": "file", "value": "src/footnoteworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/footnoteworks/report.py", "report.py"], "kind": "file", "value": "src/footnotewo...
[{"answers": ["src/footnoteworks/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-s1-00977
tariffkit
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_batch ignores the cap, so late attempts wait for minutes. It lives in src/tariffkit/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 tariffkit layout. [t...
47,248
2,362
104
111
false
synth:1724499241226080238@1.19907
synth
[{"aliases": ["src/tariffkit/cli.py", "cli.py"], "kind": "file", "value": "src/tariffkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/tariffkit/store.py", "store.py"], "kind": "file", "value": "src/tariffkit/store.py"}, {"aliases": ["data/c...
[{"answers": ["src/tariffkit/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 rec...
ctx_compress-s1-00977-twin
tariffkit
ctx_compress-s1-00977
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_batch ignores the cap, so late attempts wait for minutes. It lives in src/tariffkit/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 tariffkit layout. [t...
32,273
1,613
61
111
false
synth:1724499241226080238@1.19907
synth
[{"aliases": ["src/tariffkit/cli.py", "cli.py"], "kind": "file", "value": "src/tariffkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/tariffkit/store.py", "store.py"], "kind": "file", "value": "src/tariffkit/store.py"}, {"aliases": ["data/c...
[{"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/tariffkit/cli.py"], "ki...
ctx_compress-s1-00978
windowdesk
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_calendar averages one value too many once the window is full. It lives in src/windowdesk/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 windowdesk layout. [tool_us...
75,704
3,072
97
112
false
synth:7755907761360380564@1.49303
synth
[{"aliases": ["src/windowdesk/io.py", "io.py"], "kind": "file", "value": "src/windowdesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/windowdesk/schema.py", "schema.py"], "kind": "file", "value": "src/windowdesk/schema.py"}, {"aliases": ["src/w...
[{"answers": ["src/windowdesk/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-s1-00978-twin
windowdesk
ctx_compress-s1-00978
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_calendar averages one value too many once the window is full. It lives in src/windowdesk/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 windowdesk layout. [tool_us...
44,337
2,216
59
112
false
synth:7755907761360380564@1.49303
synth
[{"aliases": ["src/windowdesk/io.py", "io.py"], "kind": "file", "value": "src/windowdesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/windowdesk/schema.py", "schema.py"], "kind": "file", "value": "src/windowdesk/schema.py"}, {"aliases": ["src/w...
[{"answers": ["src/windowdesk/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-s1-00979
snippetkit
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_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/snippetkit/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 sn...
53,459
2,672
91
100
false
synth:3394589204761659122@1
synth
[{"aliases": ["src/snippetkit/report.py", "report.py"], "kind": "file", "value": "src/snippetkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_1.csv", "snippet_1.csv"], "kind": "file", "value": "data/snippet_1.csv"}, {"a...
[{"answers": ["src/snippetkit/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-s1-00979-twin
snippetkit
ctx_compress-s1-00979
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_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/snippetkit/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 sn...
33,431
1,671
57
100
false
synth:3394589204761659122@1
synth
[{"aliases": ["src/snippetkit/report.py", "report.py"], "kind": "file", "value": "src/snippetkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_1.csv", "snippet_1.csv"], "kind": "file", "value": "data/snippet_1.csv"}, {"a...
[{"answers": ["src/snippetkit/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-s1-00980
draftflow
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] renumber_index_window raises KeyError for a missing key instead of returning the default. It lives in src/draftflow/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 draftf...
37,303
1,865
65
72
false
synth:3345599317620221901@1
synth
[{"aliases": ["src/draftflow/cli.py", "cli.py"], "kind": "file", "value": "src/draftflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/draftflow/tables.py", "tables.py"], "kind": "file", "value": "src/draftflow/tables.py"}, {"aliases": ["dat...
[{"answers": ["src/draftflow/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 rec...
ctx_compress-s1-00980-twin
draftflow
ctx_compress-s1-00980
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] renumber_index_window raises KeyError for a missing key instead of returning the default. It lives in src/draftflow/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 draftf...
20,153
1,007
39
72
false
synth:3345599317620221901@1
synth
[{"aliases": ["src/draftflow/cli.py", "cli.py"], "kind": "file", "value": "src/draftflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/draftflow/tables.py", "tables.py"], "kind": "file", "value": "src/draftflow/tables.py"}, {"aliases": ["dat...
[{"answers": ["src/draftflow/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 rec...
ctx_compress-s1-00981
snippetflow
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_slug double counts carried amounts when the period closes on a weekend. It lives in src/snippetflow/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 snippetflow ...
32,150
1,607
57
72
false
synth:1871595343243756138@1
synth
[{"aliases": ["src/snippetflow/cli.py", "cli.py"], "kind": "file", "value": "src/snippetflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/snippetflow/schema.py", "schema.py"], "kind": "file", "value": "src/snippetflow/schema.py"}, {"aliases...
[{"answers": ["src/snippetflow/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 r...
ctx_compress-s1-00981-twin
snippetflow
ctx_compress-s1-00981
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_slug double counts carried amounts when the period closes on a weekend. It lives in src/snippetflow/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 snippetflow ...
21,427
1,071
35
72
false
synth:1871595343243756138@1
synth
[{"aliases": ["src/snippetflow/cli.py", "cli.py"], "kind": "file", "value": "src/snippetflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/snippetflow/schema.py", "schema.py"], "kind": "file", "value": "src/snippetflow/schema.py"}, {"aliases...
[{"answers": ["src/snippetflow/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 r...
ctx_compress-s1-00982
grantkit
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_policy_strict returns one item too few on every page. It lives in src/grantkit/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 grantkit layout. [tool_use kin...
61,620
3,072
98
109
false
synth:188104045829563265@1.1136
synth
[{"aliases": ["src/grantkit/report.py", "report.py"], "kind": "file", "value": "src/grantkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/grantkit/format.py", "format.py"], "kind": "file", "value": "src/grantkit/format.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": ["report.py", "src/grantkit/report.py"]...
ctx_compress-s1-00982-twin
grantkit
ctx_compress-s1-00982
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_policy_strict returns one item too few on every page. It lives in src/grantkit/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 grantkit layout. [tool_use kin...
34,919
1,745
56
109
false
synth:188104045829563265@1.1136
synth
[{"aliases": ["src/grantkit/report.py", "report.py"], "kind": "file", "value": "src/grantkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/grantkit/format.py", "format.py"], "kind": "file", "value": "src/grantkit/format.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": ["format.py", "src/grantkit/format.py"]...
ctx_compress-s1-00983
cadencedesk
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_shift rejects a range where lo equals hi, which is a valid single point range. It lives in src/cadencedesk/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 ca...
29,544
1,477
47
58
false
synth:7865811287816942129@1
synth
[{"aliases": ["src/cadencedesk/compat.py", "compat.py"], "kind": "file", "value": "src/cadencedesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/cadencedesk/rules.py", "rules.py"], "kind": "file", "value": "src/cadencedesk/rules....
[{"answers": ["src/cadencedesk/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 ...
ctx_compress-s1-00983-twin
cadencedesk
ctx_compress-s1-00983
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_shift rejects a range where lo equals hi, which is a valid single point range. It lives in src/cadencedesk/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 ca...
12,995
1,000
29
58
false
synth:7865811287816942129@1
synth
[{"aliases": ["src/cadencedesk/compat.py", "compat.py"], "kind": "file", "value": "src/cadencedesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/cadencedesk/rules.py", "rules.py"], "kind": "file", "value": "src/cadencedesk/rules....
[{"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_report.py", "tests/test_report.p...
ctx_compress-s1-00984
rosterforge
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_cadence_window raises ShiftShapeError for bytes input in strict mode instead of decoding it. It lives in src/rosterforge/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 ma...
69,515
3,072
106
127
false
synth:28233809190482389@1.88735
synth
[{"aliases": ["src/rosterforge/store.py", "store.py"], "kind": "file", "value": "src/rosterforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/rosterforge/core.py", "core.py"], "kind": "file", "value": "src/rosterforge/core.py"}, {"a...
[{"answers": ["src/rosterforge/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 mo...
ctx_compress-s1-00984-twin
rosterforge
ctx_compress-s1-00984
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_cadence_window raises ShiftShapeError for bytes input in strict mode instead of decoding it. It lives in src/rosterforge/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 ma...
33,977
1,698
66
127
false
synth:28233809190482389@1.88735
synth
[{"aliases": ["src/rosterforge/store.py", "store.py"], "kind": "file", "value": "src/rosterforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/rosterforge/core.py", "core.py"], "kind": "file", "value": "src/rosterforge/core.py"}, {"a...
[{"answers": ["src/rosterforge/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 mo...
ctx_compress-s1-00985
principalledger
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 raises KeyError for a missing key instead of returning the default. It lives in src/principalledger/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 principalle...
72,789
3,072
112
127
false
synth:9197748797849099503@2.32785
synth
[{"aliases": ["src/principalledger/io.py", "io.py"], "kind": "file", "value": "src/principalledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/principalledger/report.py", "report.py"], "kind": "file", "value": "src/principalledger/report.py"},...
[{"answers": ["src/principalledger/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...
ctx_compress-s1-00985-twin
principalledger
ctx_compress-s1-00985
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 raises KeyError for a missing key instead of returning the default. It lives in src/principalledger/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 principalle...
48,406
2,420
71
127
false
synth:9197748797849099503@2.32785
synth
[{"aliases": ["src/principalledger/io.py", "io.py"], "kind": "file", "value": "src/principalledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/principalledger/report.py", "report.py"], "kind": "file", "value": "src/principalledger/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": ["data/session_8.csv", "session_8.csv"]...
ctx_compress-s1-00986
headingkit
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] collate_section raises RevisionDriftError for bytes input in strict mode instead of decoding it. It lives in src/headingkit/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...
79,392
3,072
106
127
false
synth:8845422339152100785@2.03207
synth
[{"aliases": ["src/headingkit/store.py", "store.py"], "kind": "file", "value": "src/headingkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/headingkit/tables.py", "tables.py"], "kind": "file", "value": "src/headingkit/tables.py"}, {...
[{"answers": ["src/headingkit/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-s1-00986-twin
headingkit
ctx_compress-s1-00986
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] collate_section raises RevisionDriftError for bytes input in strict mode instead of decoding it. It lives in src/headingkit/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...
55,227
2,761
76
127
false
synth:8845422339152100785@2.03207
synth
[{"aliases": ["src/headingkit/store.py", "store.py"], "kind": "file", "value": "src/headingkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/headingkit/tables.py", "tables.py"], "kind": "file", "value": "src/headingkit/tables.py"}, {...
[{"answers": ["src/headingkit/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-s1-00987
receiptforge
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] allocate_ledger_page rejects a range where lo equals hi, which is a valid single point range. It lives in src/receiptforge/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...
67,809
3,072
100
121
false
synth:8359373077651863868@1.82445
synth
[{"aliases": ["src/receiptforge/rules.py", "rules.py"], "kind": "file", "value": "src/receiptforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/receiptforge/tables.py", "tables.py"], "kind": "file", "value": "src/receiptforge/tables...
[{"answers": ["src/receiptforge/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 m...
ctx_compress-s1-00987-twin
receiptforge
ctx_compress-s1-00987
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] allocate_ledger_page rejects a range where lo equals hi, which is a valid single point range. It lives in src/receiptforge/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...
42,611
2,130
64
121
false
synth:8359373077651863868@1.82445
synth
[{"aliases": ["src/receiptforge/rules.py", "rules.py"], "kind": "file", "value": "src/receiptforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/receiptforge/tables.py", "tables.py"], "kind": "file", "value": "src/receiptforge/tables...
[{"answers": ["src/receiptforge/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 m...
ctx_compress-s1-00988
footnotekit
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] collate_index keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/footnotekit/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 footnotekit la...
59,924
2,996
85
94
false
synth:6866604223937657472@1
synth
[{"aliases": ["src/footnotekit/core.py", "core.py"], "kind": "file", "value": "src/footnotekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/footnotekit/report.py", "report.py"], "kind": "file", "value": "src/footnotekit/report.py"}, {"a...
[{"answers": ["src/footnotekit/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...
ctx_compress-s1-00988-twin
footnotekit
ctx_compress-s1-00988
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] collate_index keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/footnotekit/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 footnotekit la...
36,855
1,842
53
94
false
synth:6866604223937657472@1
synth
[{"aliases": ["src/footnotekit/core.py", "core.py"], "kind": "file", "value": "src/footnotekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/footnotekit/report.py", "report.py"], "kind": "file", "value": "src/footnotekit/report.py"}, {"a...
[{"answers": ["src/footnotekit/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...
ctx_compress-s1-00989
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] verify_grant returns one item too few on every page. It lives in src/claimflow/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 claimflow layout. [tool_use kind=sear...
67,287
3,072
118
137
false
synth:7603848503445276785@1.8033
synth
[{"aliases": ["src/claimflow/limits.py", "limits.py"], "kind": "file", "value": "src/claimflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/policy_7.csv", "policy_7.csv"], "kind": "file", "value": "data/policy_7.csv"}, {"aliase...
[{"answers": ["src/claimflow/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-s1-00989-twin
claimflow
ctx_compress-s1-00989
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_grant returns one item too few on every page. It lives in src/claimflow/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 claimflow layout. [tool_use kind=sear...
33,807
1,690
63
137
false
synth:7603848503445276785@1.8033
synth
[{"aliases": ["src/claimflow/limits.py", "limits.py"], "kind": "file", "value": "src/claimflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/policy_7.csv", "policy_7.csv"], "kind": "file", "value": "data/policy_7.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/token_9.csv", "token_9.csv"], "k...
ctx_compress-s1-00990
grantledger
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_principal_all misses keys that differ only in case. It lives in src/grantledger/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 grantledger layout. [tool_use kind=...
50,759
2,537
78
85
false
synth:7833316961089383572@1
synth
[{"aliases": ["src/grantledger/cli.py", "cli.py"], "kind": "file", "value": "src/grantledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/policy_1.csv", "policy_1.csv"], "kind": "file", "value": "data/policy_1.csv"}, {"aliases": ["data/ses...
[{"answers": ["src/grantledger/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 r...
ctx_compress-s1-00990-twin
grantledger
ctx_compress-s1-00990
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_principal_all misses keys that differ only in case. It lives in src/grantledger/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 grantledger layout. [tool_use kind=...
35,437
1,771
50
85
false
synth:7833316961089383572@1
synth
[{"aliases": ["src/grantledger/cli.py", "cli.py"], "kind": "file", "value": "src/grantledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/policy_1.csv", "policy_1.csv"], "kind": "file", "value": "data/policy_1.csv"}, {"aliases": ["data/ses...
[{"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/policy_1.csv", "policy_1.csv"], ...
ctx_compress-s1-00991
creditkit
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_credit produces doubled dashes for consecutive separators. It lives in src/creditkit/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 creditkit layout. [tool_...
27,994
1,399
45
52
false
synth:1608043436203337758@1
synth
[{"aliases": ["src/creditkit/report.py", "report.py"], "kind": "file", "value": "src/creditkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/creditkit/tables.py", "tables.py"], "kind": "file", "value": "src/creditkit/tables.py"},...
[{"answers": ["src/creditkit/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-s1-00991-twin
creditkit
ctx_compress-s1-00991
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_credit produces doubled dashes for consecutive separators. It lives in src/creditkit/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 creditkit layout. [tool_...
14,574
1,000
31
52
false
synth:1608043436203337758@1
synth
[{"aliases": ["src/creditkit/report.py", "report.py"], "kind": "file", "value": "src/creditkit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/creditkit/tables.py", "tables.py"], "kind": "file", "value": "src/creditkit/tables.py"},...
[{"answers": ["src/creditkit/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-s1-00992
noncedesk
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_principal misses keys that differ only in case. It lives in src/noncedesk/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 noncedesk layout. [tool_use kind=se...
55,001
2,750
102
111
false
synth:4249947448239676399@1.2331
synth
[{"aliases": ["src/noncedesk/limits.py", "limits.py"], "kind": "file", "value": "src/noncedesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/noncedesk/schema.py", "schema.py"], "kind": "file", "value": "src/noncedesk/schema.py"},...
[{"answers": ["src/noncedesk/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-s1-00992-twin
noncedesk
ctx_compress-s1-00992
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_principal misses keys that differ only in case. It lives in src/noncedesk/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 noncedesk layout. [tool_use kind=se...
32,996
1,649
54
111
false
synth:4249947448239676399@1.2331
synth
[{"aliases": ["src/noncedesk/limits.py", "limits.py"], "kind": "file", "value": "src/noncedesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/noncedesk/schema.py", "schema.py"], "kind": "file", "value": "src/noncedesk/schema.py"},...
[{"answers": ["src/noncedesk/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-s1-00993
palletforge
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_pallet_strict averages one value too many once the window is full. It lives in src/palletforge/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 palletforge layo...
100,955
3,072
129
138
false
synth:6519690134460978110@2.22892
synth
[{"aliases": ["src/palletforge/core.py", "core.py"], "kind": "file", "value": "src/palletforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/palletforge/store.py", "store.py"], "kind": "file", "value": "src/palletforge/store.py"}, {"alia...
[{"answers": ["src/palletforge/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...
ctx_compress-s1-00993-twin
palletforge
ctx_compress-s1-00993
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_pallet_strict averages one value too many once the window is full. It lives in src/palletforge/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 palletforge layo...
51,575
2,578
67
138
false
synth:6519690134460978110@2.22892
synth
[{"aliases": ["src/palletforge/core.py", "core.py"], "kind": "file", "value": "src/palletforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/palletforge/store.py", "store.py"], "kind": "file", "value": "src/palletforge/store.py"}, {"alia...
[{"answers": ["src/palletforge/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-s1-00994
receiptforge
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_payout_safe grows linearly instead of doubling. It lives in src/receiptforge/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 receiptforge layout. [tool_use kind=sea...
79,114
3,072
111
140
false
synth:8105524044906131943@2.30187
synth
[{"aliases": ["src/receiptforge/core.py", "core.py"], "kind": "file", "value": "src/receiptforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/receiptforge/tables.py", "tables.py"], "kind": "file", "value": "src/receiptforge/tables.py"},...
[{"answers": ["src/receiptforge/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-s1-00994-twin
receiptforge
ctx_compress-s1-00994
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_payout_safe grows linearly instead of doubling. It lives in src/receiptforge/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 receiptforge layout. [tool_use kind=sea...
50,899
2,544
73
140
false
synth:8105524044906131943@2.30187
synth
[{"aliases": ["src/receiptforge/core.py", "core.py"], "kind": "file", "value": "src/receiptforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/receiptforge/tables.py", "tables.py"], "kind": "file", "value": "src/receiptforge/tables.py"},...
[{"answers": ["src/receiptforge/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 mos...
ctx_compress-s1-00995
tariffflow
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_receipt rejects a range where lo equals hi, which is a valid single point range. It lives in src/tariffflow/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 tar...
63,946
3,072
102
115
false
synth:1075279441638032472@1.33868
synth
[{"aliases": ["src/tariffflow/format.py", "format.py"], "kind": "file", "value": "src/tariffflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tariffflow/tables.py", "tables.py"], "kind": "file", "value": "src/tariffflow/tables.p...
[{"answers": ["src/tariffflow/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 m...
ctx_compress-s1-00995-twin
tariffflow
ctx_compress-s1-00995
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_receipt rejects a range where lo equals hi, which is a valid single point range. It lives in src/tariffflow/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 tar...
49,966
2,498
68
115
false
synth:1075279441638032472@1.33868
synth
[{"aliases": ["src/tariffflow/format.py", "format.py"], "kind": "file", "value": "src/tariffflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tariffflow/tables.py", "tables.py"], "kind": "file", "value": "src/tariffflow/tables.p...
[{"answers": ["src/tariffflow/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 m...
ctx_compress-s1-00996
waypointdesk
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_corridor_once raises WaypointStateError for bytes input in strict mode instead of decoding it. It lives in src/waypointdesk/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 star...
95,029
3,072
114
133
false
synth:5997623464133607911@2.05547
synth
[{"aliases": ["src/waypointdesk/format.py", "format.py"], "kind": "file", "value": "src/waypointdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/zone_1.csv", "zone_1.csv"], "kind": "file", "value": "data/zone_1.csv"}, {"aliase...
[{"answers": ["src/waypointdesk/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-s1-00996-twin
waypointdesk
ctx_compress-s1-00996
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_corridor_once raises WaypointStateError for bytes input in strict mode instead of decoding it. It lives in src/waypointdesk/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 star...
67,353
3,072
78
133
false
synth:5997623464133607911@2.05547
synth
[{"aliases": ["src/waypointdesk/format.py", "format.py"], "kind": "file", "value": "src/waypointdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/zone_1.csv", "zone_1.csv"], "kind": "file", "value": "data/zone_1.csv"}, {"aliase...
[{"answers": ["src/waypointdesk/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-s1-00997
refundflow
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] allocate_ledger_batch skips the first page; page 1 returns what should be page 2. It lives in src/refundflow/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 refundf...
99,425
3,072
131
148
false
synth:7216714108587463421@2.3298
synth
[{"aliases": ["src/refundflow/format.py", "format.py"], "kind": "file", "value": "src/refundflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/ledger_9.csv", "ledger_9.csv"], "kind": "file", "value": "data/ledger_9.csv"}, {"alia...
[{"answers": ["src/refundflow/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 m...
ctx_compress-s1-00997-twin
refundflow
ctx_compress-s1-00997
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] allocate_ledger_batch skips the first page; page 1 returns what should be page 2. It lives in src/refundflow/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 refundf...
59,247
2,962
77
148
false
synth:7216714108587463421@2.3298
synth
[{"aliases": ["src/refundflow/format.py", "format.py"], "kind": "file", "value": "src/refundflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/ledger_9.csv", "ledger_9.csv"], "kind": "file", "value": "data/ledger_9.csv"}, {"alia...
[{"answers": ["src/refundflow/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 m...
ctx_compress-s1-00998
fixforge
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] interpolate_bearing_page raises KeyError for a missing key instead of returning the default. It lives in src/fixforge/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 th...
33,432
1,671
48
59
false
synth:2284813560179803916@1
synth
[{"aliases": ["src/fixforge/format.py", "format.py"], "kind": "file", "value": "src/fixforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/fix_3.csv", "fix_3.csv"], "kind": "file", "value": "data/fix_3.csv"}, {"aliases": ["data/...
[{"answers": ["src/fixforge/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 mos...
ctx_compress-s1-00998-twin
fixforge
ctx_compress-s1-00998
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] interpolate_bearing_page raises KeyError for a missing key instead of returning the default. It lives in src/fixforge/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 th...
21,138
1,056
32
59
false
synth:2284813560179803916@1
synth
[{"aliases": ["src/fixforge/format.py", "format.py"], "kind": "file", "value": "src/fixforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/fix_3.csv", "fix_3.csv"], "kind": "file", "value": "data/fix_3.csv"}, {"aliases": ["data/...
[{"answers": ["src/fixforge/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 mos...
ctx_compress-s1-00999
slotworks
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_window double counts carried amounts when the period closes on a weekend. It lives in src/slotworks/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 slotworks...
38,659
1,932
73
78
false
synth:3606776970041756925@1
synth
[{"aliases": ["src/slotworks/format.py", "format.py"], "kind": "file", "value": "src/slotworks/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slotworks/limits.py", "limits.py"], "kind": "file", "value": "src/slotworks/limits.py"},...
[{"answers": ["src/slotworks/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-s1-00999-twin
slotworks
ctx_compress-s1-00999
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_window double counts carried amounts when the period closes on a weekend. It lives in src/slotworks/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 slotworks...
22,368
1,118
45
78
false
synth:3606776970041756925@1
synth
[{"aliases": ["src/slotworks/format.py", "format.py"], "kind": "file", "value": "src/slotworks/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slotworks/limits.py", "limits.py"], "kind": "file", "value": "src/slotworks/limits.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": ["limits.py", "src/slotworks/limits.py"...
ctx_compress-s1-01000
fixflow
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_fix raises TileDriftError for bytes input in strict mode instead of decoding it. It lives in src/fixflow/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 fix...
74,164
3,072
123
132
false
synth:7337874933463744741@1.74422
synth
[{"aliases": ["src/fixflow/report.py", "report.py"], "kind": "file", "value": "src/fixflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/fixflow/schema.py", "schema.py"], "kind": "file", "value": "src/fixflow/schema.py"}, {"alias...
[{"answers": ["src/fixflow/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-s1-01000-twin
fixflow
ctx_compress-s1-01000
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_fix raises TileDriftError for bytes input in strict mode instead of decoding it. It lives in src/fixflow/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 fix...
35,930
1,796
69
132
false
synth:7337874933463744741@1.74422
synth
[{"aliases": ["src/fixflow/report.py", "report.py"], "kind": "file", "value": "src/fixflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/fixflow/schema.py", "schema.py"], "kind": "file", "value": "src/fixflow/schema.py"}, {"alias...
[{"answers": ["src/fixflow/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 most...
ctx_compress-s1-01001
sessionflow
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 returns one item too few on every page. It lives in src/sessionflow/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 sessionflow layout. [tool_use kin...
104,579
3,072
130
149
false
synth:6820393141058472773@2.31055
synth
[{"aliases": ["src/sessionflow/rules.py", "rules.py"], "kind": "file", "value": "src/sessionflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/grant_3.csv", "grant_3.csv"], "kind": "file", "value": "data/grant_3.csv"}, {"aliases": [...
[{"answers": ["src/sessionflow/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 mo...
ctx_compress-s1-01001-twin
sessionflow
ctx_compress-s1-01001
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 returns one item too few on every page. It lives in src/sessionflow/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 sessionflow layout. [tool_use kin...
51,288
2,564
68
149
false
synth:6820393141058472773@2.31055
synth
[{"aliases": ["src/sessionflow/rules.py", "rules.py"], "kind": "file", "value": "src/sessionflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/grant_3.csv", "grant_3.csv"], "kind": "file", "value": "data/grant_3.csv"}, {"aliases": [...
[{"answers": ["src/sessionflow/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 mo...
ctx_compress-s1-01002
lotforge
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 raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/lotforge/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 lotfor...
56,948
2,847
96
111
false
synth:4620958036657880622@1.05808
synth
[{"aliases": ["src/lotforge/report.py", "report.py"], "kind": "file", "value": "src/lotforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/lotforge/core.py", "core.py"], "kind": "file", "value": "src/lotforge/core.py"}, {"aliases...
[{"answers": ["src/lotforge/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 mos...
ctx_compress-s1-01002-twin
lotforge
ctx_compress-s1-01002
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 raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/lotforge/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 lotfor...
29,689
1,484
52
111
false
synth:4620958036657880622@1.05808
synth
[{"aliases": ["src/lotforge/report.py", "report.py"], "kind": "file", "value": "src/lotforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/lotforge/core.py", "core.py"], "kind": "file", "value": "src/lotforge/core.py"}, {"aliases...
[{"answers": ["src/lotforge/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 mos...
ctx_compress-s1-01003
palletworks
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 raises SkuShapeError for bytes input in strict mode instead of decoding it. It lives in src/palletworks/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 p...
57,222
2,861
85
94
false
synth:8556840716195631224@1
synth
[{"aliases": ["src/palletworks/rules.py", "rules.py"], "kind": "file", "value": "src/palletworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/palletworks/schema.py", "schema.py"], "kind": "file", "value": "src/palletworks/schema.py"...
[{"answers": ["src/palletworks/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 mo...
ctx_compress-s1-01003-twin
palletworks
ctx_compress-s1-01003
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 raises SkuShapeError for bytes input in strict mode instead of decoding it. It lives in src/palletworks/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 p...
24,166
1,208
49
94
false
synth:8556840716195631224@1
synth
[{"aliases": ["src/palletworks/rules.py", "rules.py"], "kind": "file", "value": "src/palletworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/palletworks/schema.py", "schema.py"], "kind": "file", "value": "src/palletworks/schema.py"...
[{"answers": ["src/palletworks/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 mo...
ctx_compress-s1-01004
windowflow
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_shift returns one item too few on every page. It lives in src/windowflow/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 windowflow layout. [tool_use kind=search ...
86,958
3,072
110
137
false
synth:8604919076506325648@1.86135
synth
[{"aliases": ["src/windowflow/core.py", "core.py"], "kind": "file", "value": "src/windowflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/cadence_3.csv", "cadence_3.csv"], "kind": "file", "value": "data/cadence_3.csv"}, {"aliases": ["s...
[{"answers": ["src/windowflow/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 ...
ctx_compress-s1-01004-twin
windowflow
ctx_compress-s1-01004
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_shift returns one item too few on every page. It lives in src/windowflow/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 windowflow layout. [tool_use kind=search ...
63,122
3,072
78
137
false
synth:8604919076506325648@1.86135
synth
[{"aliases": ["src/windowflow/core.py", "core.py"], "kind": "file", "value": "src/windowflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/cadence_3.csv", "cadence_3.csv"], "kind": "file", "value": "data/cadence_3.csv"}, {"aliases": ["s...
[{"answers": ["src/windowflow/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 ...
ctx_compress-s1-01005
tokendesk
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_policy_batch skips the first page; page 1 returns what should be page 2. It lives in src/tokendesk/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 tokendesk ...
67,546
3,072
100
119
false
synth:8777427019572643191@1.6746
synth
[{"aliases": ["src/tokendesk/limits.py", "limits.py"], "kind": "file", "value": "src/tokendesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/scope_4.csv", "scope_4.csv"], "kind": "file", "value": "data/scope_4.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": ["cli.py", "src/tokendesk/cli.py"], "ki...