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-s1-00000
snippetdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_slug_window produces doubled dashes for consecutive separators. It lives in src/snippetdesk/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 snippetdesk layout....
31,317
1,565
56
67
false
synth:5995469358269906430@1
synth
[{"aliases": ["src/snippetdesk/limits.py", "limits.py"], "kind": "file", "value": "src/snippetdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/chunk_1.csv", "chunk_1.csv"], "kind": "file", "value": "data/chunk_1.csv"}, {"alias...
[{"answers": ["src/snippetdesk/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 ...
ctx_compress-s1-00000-twin
snippetdesk
ctx_compress-s1-00000
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_slug_window produces doubled dashes for consecutive separators. It lives in src/snippetdesk/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 snippetdesk layout....
18,066
1,000
40
67
false
synth:5995469358269906430@1
synth
[{"aliases": ["src/snippetdesk/limits.py", "limits.py"], "kind": "file", "value": "src/snippetdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/chunk_1.csv", "chunk_1.csv"], "kind": "file", "value": "data/chunk_1.csv"}, {"alias...
[{"answers": ["src/snippetdesk/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 ...
ctx_compress-s1-00001
draftledger
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] fold_clause_window rejects a range where lo equals hi, which is a valid single point range. It lives in src/draftledger/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 ...
66,549
3,072
84
103
false
synth:7735715960199495141@2.32595
synth
[{"aliases": ["src/draftledger/core.py", "core.py"], "kind": "file", "value": "src/draftledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/draftledger/limits.py", "limits.py"], "kind": "file", "value": "src/draftledger/limits.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": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00001-twin
draftledger
ctx_compress-s1-00001
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] fold_clause_window rejects a range where lo equals hi, which is a valid single point range. It lives in src/draftledger/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 ...
40,704
2,035
55
103
false
synth:7735715960199495141@2.32595
synth
[{"aliases": ["src/draftledger/core.py", "core.py"], "kind": "file", "value": "src/draftledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/draftledger/limits.py", "limits.py"], "kind": "file", "value": "src/draftledger/limits.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": ["audit.py", "src/draftledger/audit.py"...
ctx_compress-s1-00002
spanledger
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] smooth_counter_all raises ValueError when a value itself contains an equals sign. It lives in src/spanledger/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 spanledger layo...
65,234
3,072
99
124
false
synth:3719260088301377541@2.30043
synth
[{"aliases": ["src/spanledger/io.py", "io.py"], "kind": "file", "value": "src/spanledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/probe_2.csv", "probe_2.csv"], "kind": "file", "value": "data/probe_2.csv"}, {"aliases": ["src/spanledger/comp...
[{"answers": ["src/spanledger/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-s1-00002-twin
spanledger
ctx_compress-s1-00002
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] smooth_counter_all raises ValueError when a value itself contains an equals sign. It lives in src/spanledger/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 spanledger layo...
50,824
2,541
71
124
false
synth:3719260088301377541@2.30043
synth
[{"aliases": ["src/spanledger/io.py", "io.py"], "kind": "file", "value": "src/spanledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/probe_2.csv", "probe_2.csv"], "kind": "file", "value": "data/probe_2.csv"}, {"aliases": ["src/spanledger/comp...
[{"answers": ["src/spanledger/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-s1-00003
sessionledger
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_grant lets defaults overwrite explicit overrides. It lives in src/sessionledger/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 sessionledger layout. [tool_use...
41,326
2,066
87
94
false
synth:4826020194796927585@1
synth
[{"aliases": ["src/sessionledger/audit.py", "audit.py"], "kind": "file", "value": "src/sessionledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sessionledger/tables.py", "tables.py"], "kind": "file", "value": "src/sessionledger/ta...
[{"answers": ["src/sessionledger/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 ...
ctx_compress-s1-00003-twin
sessionledger
ctx_compress-s1-00003
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_grant lets defaults overwrite explicit overrides. It lives in src/sessionledger/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 sessionledger layout. [tool_use...
26,547
1,327
49
94
false
synth:4826020194796927585@1
synth
[{"aliases": ["src/sessionledger/audit.py", "audit.py"], "kind": "file", "value": "src/sessionledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sessionledger/tables.py", "tables.py"], "kind": "file", "value": "src/sessionledger/ta...
[{"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/principal_3.csv", "principal_3.c...
ctx_compress-s1-00004
sessiondesk
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_grant raises KeyError for a missing key instead of returning the default. It lives in src/sessiondesk/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 sessionde...
56,850
2,842
96
103
false
synth:2636109939138424196@1.22435
synth
[{"aliases": ["src/sessiondesk/audit.py", "audit.py"], "kind": "file", "value": "src/sessiondesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sessiondesk/format.py", "format.py"], "kind": "file", "value": "src/sessiondesk/format.py"...
[{"answers": ["src/sessiondesk/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 mo...
ctx_compress-s1-00004-twin
sessiondesk
ctx_compress-s1-00004
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_grant raises KeyError for a missing key instead of returning the default. It lives in src/sessiondesk/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 sessionde...
31,893
1,594
57
103
false
synth:2636109939138424196@1.22435
synth
[{"aliases": ["src/sessiondesk/audit.py", "audit.py"], "kind": "file", "value": "src/sessiondesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sessiondesk/format.py", "format.py"], "kind": "file", "value": "src/sessiondesk/format.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": ["audit.py", "src/sessiondesk/audit.py"...
ctx_compress-s1-00005
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] roll_credit raises ValueError when a value itself contains an equals sign. It lives in src/creditkit/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 creditkit layou...
24,165
1,208
58
65
false
synth:3689816042249187310@1
synth
[{"aliases": ["src/creditkit/format.py", "format.py"], "kind": "file", "value": "src/creditkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/invoice_1.csv", "invoice_1.csv"], "kind": "file", "value": "data/invoice_1.csv"}, {"ali...
[{"answers": ["src/creditkit/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-s1-00005-twin
creditkit
ctx_compress-s1-00005
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_credit raises ValueError when a value itself contains an equals sign. It lives in src/creditkit/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 creditkit layou...
16,156
1,000
32
65
false
synth:3689816042249187310@1
synth
[{"aliases": ["src/creditkit/format.py", "format.py"], "kind": "file", "value": "src/creditkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/invoice_1.csv", "invoice_1.csv"], "kind": "file", "value": "data/invoice_1.csv"}, {"ali...
[{"answers": ["src/creditkit/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-s1-00006
captionworks
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_glyph double counts carried amounts when the period closes on a weekend. It lives in src/captionworks/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 captio...
44,555
2,227
84
103
false
synth:3537910650999356050@1.42447
synth
[{"aliases": ["src/captionworks/audit.py", "audit.py"], "kind": "file", "value": "src/captionworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/captionworks/limits.py", "limits.py"], "kind": "file", "value": "src/captionworks/limits...
[{"answers": ["src/captionworks/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 m...
ctx_compress-s1-00006-twin
captionworks
ctx_compress-s1-00006
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_glyph double counts carried amounts when the period closes on a weekend. It lives in src/captionworks/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 captio...
31,257
1,562
53
103
false
synth:3537910650999356050@1.42447
synth
[{"aliases": ["src/captionworks/audit.py", "audit.py"], "kind": "file", "value": "src/captionworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/captionworks/limits.py", "limits.py"], "kind": "file", "value": "src/captionworks/limits...
[{"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_core.py", "tests/test_core.py"],...
ctx_compress-s1-00007
receiptflow
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_tariff rejects a range where lo equals hi, which is a valid single point range. It lives in src/receiptflow/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 ...
37,073
1,853
60
75
false
synth:2211135390217156495@1
synth
[{"aliases": ["src/receiptflow/store.py", "store.py"], "kind": "file", "value": "src/receiptflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/receiptflow/io.py", "io.py"], "kind": "file", "value": "src/receiptflow/io.py"}, {"aliases...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_compat.py", "tests/test_compat.p...
ctx_compress-s1-00007-twin
receiptflow
ctx_compress-s1-00007
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_tariff rejects a range where lo equals hi, which is a valid single point range. It lives in src/receiptflow/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 ...
22,591
1,129
40
75
false
synth:2211135390217156495@1
synth
[{"aliases": ["src/receiptflow/store.py", "store.py"], "kind": "file", "value": "src/receiptflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/receiptflow/io.py", "io.py"], "kind": "file", "value": "src/receiptflow/io.py"}, {"aliases...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00008
glyphkit
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_stem produces doubled dashes for consecutive separators. It lives in src/glyphkit/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 glyphkit layout. [tool_us...
63,273
3,072
88
97
false
synth:3040212208081554271@1.0628
synth
[{"aliases": ["src/glyphkit/compat.py", "compat.py"], "kind": "file", "value": "src/glyphkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/glyphkit/rules.py", "rules.py"], "kind": "file", "value": "src/glyphkit/rules.py"}, {"alia...
[{"answers": ["src/glyphkit/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 mos...
ctx_compress-s1-00008-twin
glyphkit
ctx_compress-s1-00008
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_stem produces doubled dashes for consecutive separators. It lives in src/glyphkit/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 glyphkit layout. [tool_us...
38,944
1,947
54
97
false
synth:3040212208081554271@1.0628
synth
[{"aliases": ["src/glyphkit/compat.py", "compat.py"], "kind": "file", "value": "src/glyphkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/glyphkit/rules.py", "rules.py"], "kind": "file", "value": "src/glyphkit/rules.py"}, {"alia...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/glyphkit/store.py", "store.py"], ...
ctx_compress-s1-00009
tokenkit
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_scope raises ValueError when a value itself contains an equals sign. It lives in src/tokenkit/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 tokenkit layout. ...
59,334
2,966
82
89
false
synth:6662391810543765172@1
synth
[{"aliases": ["src/tokenkit/store.py", "store.py"], "kind": "file", "value": "src/tokenkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/nonce_3.csv", "nonce_3.csv"], "kind": "file", "value": "data/nonce_3.csv"}, {"aliases": ["src/t...
[{"answers": ["src/tokenkit/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 most ...
ctx_compress-s1-00009-twin
tokenkit
ctx_compress-s1-00009
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_scope raises ValueError when a value itself contains an equals sign. It lives in src/tokenkit/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 tokenkit layout. ...
26,013
1,300
42
89
false
synth:6662391810543765172@1
synth
[{"aliases": ["src/tokenkit/store.py", "store.py"], "kind": "file", "value": "src/tokenkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/nonce_3.csv", "nonce_3.csv"], "kind": "file", "value": "data/nonce_3.csv"}, {"aliases": ["src/t...
[{"answers": ["src/tokenkit/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 most ...
ctx_compress-s1-00010
waypointforge
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_waypoint raises KeyError for a missing key instead of returning the default. It lives in src/waypointforge/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 ...
85,040
3,072
103
114
false
synth:6466448468994320724@1.52685
synth
[{"aliases": ["src/waypointforge/report.py", "report.py"], "kind": "file", "value": "src/waypointforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/waypointforge/cli.py", "cli.py"], "kind": "file", "value": "src/waypointforge/cl...
[{"answers": ["src/waypointforge/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'...
ctx_compress-s1-00010-twin
waypointforge
ctx_compress-s1-00010
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_waypoint raises KeyError for a missing key instead of returning the default. It lives in src/waypointforge/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 ...
43,152
2,157
61
114
false
synth:6466448468994320724@1.52685
synth
[{"aliases": ["src/waypointforge/report.py", "report.py"], "kind": "file", "value": "src/waypointforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/waypointforge/cli.py", "cli.py"], "kind": "file", "value": "src/waypointforge/cl...
[{"answers": ["src/waypointforge/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'...
ctx_compress-s1-00011
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] rotate_principal returns the upper bound for every value instead of clamping into the range. It lives in src/grantkit/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 grantk...
85,007
3,072
100
119
false
synth:2281071645472042807@1.84327
synth
[{"aliases": ["src/grantkit/io.py", "io.py"], "kind": "file", "value": "src/grantkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_7.csv", "grant_7.csv"], "kind": "file", "value": "data/grant_7.csv"}, {"aliases": ["docs/changelog.md", "ch...
[{"answers": ["src/grantkit/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-s1-00011-twin
grantkit
ctx_compress-s1-00011
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 returns the upper bound for every value instead of clamping into the range. It lives in src/grantkit/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 grantk...
45,756
2,287
58
119
false
synth:2281071645472042807@1.84327
synth
[{"aliases": ["src/grantkit/io.py", "io.py"], "kind": "file", "value": "src/grantkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_7.csv", "grant_7.csv"], "kind": "file", "value": "data/grant_7.csv"}, {"aliases": ["docs/changelog.md", "ch...
[{"answers": ["src/grantkit/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-s1-00012
marginflow
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] fold_revision_once double counts carried amounts when the period closes on a weekend. It lives in src/marginflow/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 mar...
76,504
3,072
100
113
false
synth:7429451298348421847@1.69405
synth
[{"aliases": ["src/marginflow/compat.py", "compat.py"], "kind": "file", "value": "src/marginflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/revision_2.csv", "revision_2.csv"], "kind": "file", "value": "data/revision_2.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_format.py", "tests/test_format.p...
ctx_compress-s1-00012-twin
marginflow
ctx_compress-s1-00012
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] fold_revision_once double counts carried amounts when the period closes on a weekend. It lives in src/marginflow/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 mar...
35,594
1,779
52
113
false
synth:7429451298348421847@1.69405
synth
[{"aliases": ["src/marginflow/compat.py", "compat.py"], "kind": "file", "value": "src/marginflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/revision_2.csv", "revision_2.csv"], "kind": "file", "value": "data/revision_2.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/marginflow/format.py...
ctx_compress-s1-00013
gaugedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket double counts carried amounts when the period closes on a weekend. It lives in src/gaugedesk/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 gaugedesk lay...
64,069
3,072
101
128
false
synth:9118685607624819110@1.5101
synth
[{"aliases": ["src/gaugedesk/audit.py", "audit.py"], "kind": "file", "value": "src/gaugedesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/gaugedesk/format.py", "format.py"], "kind": "file", "value": "src/gaugedesk/format.py"}, {"ali...
[{"answers": ["src/gaugedesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-00013-twin
gaugedesk
ctx_compress-s1-00013
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket double counts carried amounts when the period closes on a weekend. It lives in src/gaugedesk/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 gaugedesk lay...
35,781
1,789
59
128
false
synth:9118685607624819110@1.5101
synth
[{"aliases": ["src/gaugedesk/audit.py", "audit.py"], "kind": "file", "value": "src/gaugedesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/gaugedesk/format.py", "format.py"], "kind": "file", "value": "src/gaugedesk/format.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": ["format.py", "src/gaugedesk/format.py"...
ctx_compress-s1-00014
slugledger
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_cue_once drops override values of 0, False and empty string. It lives in src/slugledger/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 slugledger layout. [tool_us...
29,141
1,457
55
60
false
synth:8538983751686835756@1
synth
[{"aliases": ["src/slugledger/io.py", "io.py"], "kind": "file", "value": "src/slugledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/stem_3.csv", "stem_3.csv"], "kind": "file", "value": "data/stem_3.csv"}, {"aliases": ["src/slugledger/compat....
[{"answers": ["src/slugledger/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-s1-00014-twin
slugledger
ctx_compress-s1-00014
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_cue_once drops override values of 0, False and empty string. It lives in src/slugledger/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 slugledger layout. [tool_us...
12,134
1,000
27
60
false
synth:8538983751686835756@1
synth
[{"aliases": ["src/slugledger/io.py", "io.py"], "kind": "file", "value": "src/slugledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/stem_3.csv", "stem_3.csv"], "kind": "file", "value": "data/stem_3.csv"}, {"aliases": ["src/slugledger/compat....
[{"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/slugledger/report.py...
ctx_compress-s1-00015
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] pick_reorder rejects a range where lo equals hi, which is a valid single point range. It lives in src/palletworks/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 pa...
69,017
3,072
109
124
false
synth:7960136413199065075@2.08637
synth
[{"aliases": ["src/palletworks/report.py", "report.py"], "kind": "file", "value": "src/palletworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/palletworks/format.py", "format.py"], "kind": "file", "value": "src/palletworks/form...
[{"answers": ["src/palletworks/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-s1-00015-twin
palletworks
ctx_compress-s1-00015
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_reorder rejects a range where lo equals hi, which is a valid single point range. It lives in src/palletworks/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 pa...
41,975
2,098
63
124
false
synth:7960136413199065075@2.08637
synth
[{"aliases": ["src/palletworks/report.py", "report.py"], "kind": "file", "value": "src/palletworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/palletworks/format.py", "format.py"], "kind": "file", "value": "src/palletworks/form...
[{"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/lot_7.csv", "lot_7.csv"], "kind"...
ctx_compress-s1-00016
sessionledger
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 grows linearly instead of doubling. It lives in src/sessionledger/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 sessionledger layout. [tool_use kind=se...
105,973
3,072
109
122
false
synth:6076815695595723977@2.25875
synth
[{"aliases": ["src/sessionledger/cli.py", "cli.py"], "kind": "file", "value": "src/sessionledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/nonce_9.csv", "nonce_9.csv"], "kind": "file", "value": "data/nonce_9.csv"}, {"aliases": ["data/gr...
[{"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/sessionledger/format...
ctx_compress-s1-00016-twin
sessionledger
ctx_compress-s1-00016
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 grows linearly instead of doubling. It lives in src/sessionledger/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 sessionledger layout. [tool_use kind=se...
57,787
2,889
55
122
false
synth:6076815695595723977@2.25875
synth
[{"aliases": ["src/sessionledger/cli.py", "cli.py"], "kind": "file", "value": "src/sessionledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/nonce_9.csv", "nonce_9.csv"], "kind": "file", "value": "data/nonce_9.csv"}, {"aliases": ["data/gr...
[{"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/principal_8.csv", "principal_8.c...
ctx_compress-s1-00017
zoneflow
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_safe lets defaults overwrite explicit overrides. It lives in src/zoneflow/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 zoneflow layout. [tool_use kin...
41,707
2,085
101
108
false
synth:3330542949785824601@1
synth
[{"aliases": ["src/zoneflow/rules.py", "rules.py"], "kind": "file", "value": "src/zoneflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/zoneflow/cli.py", "cli.py"], "kind": "file", "value": "src/zoneflow/cli.py"}, {"aliases": ["src/...
[{"answers": ["src/zoneflow/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-s1-00017-twin
zoneflow
ctx_compress-s1-00017
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_safe lets defaults overwrite explicit overrides. It lives in src/zoneflow/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 zoneflow layout. [tool_use kin...
18,995
1,000
55
108
false
synth:3330542949785824601@1
synth
[{"aliases": ["src/zoneflow/rules.py", "rules.py"], "kind": "file", "value": "src/zoneflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/zoneflow/cli.py", "cli.py"], "kind": "file", "value": "src/zoneflow/cli.py"}, {"aliases": ["src/...
[{"answers": ["src/zoneflow/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-s1-00018
counterledger
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] bin_series_once raises TraceShapeError for bytes input in strict mode instead of decoding it. It lives in src/counterledger/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 mappin...
47,341
2,367
96
105
false
synth:7825018132334369408@1
synth
[{"aliases": ["src/counterledger/audit.py", "audit.py"], "kind": "file", "value": "src/counterledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["src/counterledger/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 ...
ctx_compress-s1-00018-twin
counterledger
ctx_compress-s1-00018
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] bin_series_once raises TraceShapeError for bytes input in strict mode instead of decoding it. It lives in src/counterledger/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 mappin...
28,976
1,448
62
105
false
synth:7825018132334369408@1
synth
[{"aliases": ["src/counterledger/audit.py", "audit.py"], "kind": "file", "value": "src/counterledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_format.py", "tests/test_format.p...
ctx_compress-s1-00019
bookingflow
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 raises ShiftRangeError for bytes input in strict mode instead of decoding it. It lives in src/bookingflow/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 th...
91,859
3,072
113
122
false
synth:7645769512357244348@1.5626
synth
[{"aliases": ["src/bookingflow/compat.py", "compat.py"], "kind": "file", "value": "src/bookingflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/bookingflow/store.py", "store.py"], "kind": "file", "value": "src/bookingflow/store....
[{"answers": ["src/bookingflow/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-s1-00019-twin
bookingflow
ctx_compress-s1-00019
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 raises ShiftRangeError for bytes input in strict mode instead of decoding it. It lives in src/bookingflow/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 th...
46,599
2,329
61
122
false
synth:7645769512357244348@1.5626
synth
[{"aliases": ["src/bookingflow/compat.py", "compat.py"], "kind": "file", "value": "src/bookingflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/bookingflow/store.py", "store.py"], "kind": "file", "value": "src/bookingflow/store....
[{"answers": ["src/bookingflow/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-s1-00020
invoiceforge
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_balance lets defaults overwrite explicit overrides. It lives in src/invoiceforge/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 invoiceforge layout. [tool_use kind=...
76,982
3,072
98
111
false
synth:4021640784853194672@2.14158
synth
[{"aliases": ["src/invoiceforge/cli.py", "cli.py"], "kind": "file", "value": "src/invoiceforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/tariff_4.csv", "tariff_4.csv"], "kind": "file", "value": "data/tariff_4.csv"}, {"aliases": ["src/in...
[{"answers": ["src/invoiceforge/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-s1-00020-twin
invoiceforge
ctx_compress-s1-00020
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_balance lets defaults overwrite explicit overrides. It lives in src/invoiceforge/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 invoiceforge layout. [tool_use kind=...
49,166
2,458
57
111
false
synth:4021640784853194672@2.14158
synth
[{"aliases": ["src/invoiceforge/cli.py", "cli.py"], "kind": "file", "value": "src/invoiceforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/tariff_4.csv", "tariff_4.csv"], "kind": "file", "value": "data/tariff_4.csv"}, {"aliases": ["src/in...
[{"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/invoiceforge/cli.py"], ...
ctx_compress-s1-00021
captionflow
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] dedupe_chunk raises ValueError when a value itself contains an equals sign. It lives in src/captionflow/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 captionflow la...
56,749
2,837
96
115
false
synth:4160646128030017728@1.215
synth
[{"aliases": ["src/captionflow/audit.py", "audit.py"], "kind": "file", "value": "src/captionflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/captionflow/io.py", "io.py"], "kind": "file", "value": "src/captionflow/io.py"}, {"aliases...
[{"answers": ["src/captionflow/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 mo...
ctx_compress-s1-00021-twin
captionflow
ctx_compress-s1-00021
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] dedupe_chunk raises ValueError when a value itself contains an equals sign. It lives in src/captionflow/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 captionflow la...
37,395
1,869
66
115
false
synth:4160646128030017728@1.215
synth
[{"aliases": ["src/captionflow/audit.py", "audit.py"], "kind": "file", "value": "src/captionflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/captionflow/io.py", "io.py"], "kind": "file", "value": "src/captionflow/io.py"}, {"aliases...
[{"answers": ["src/captionflow/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 mo...
ctx_compress-s1-00022
stemworks
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_stem double counts carried amounts when the period closes on a weekend. It lives in src/stemworks/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 stemworks ...
64,162
3,072
90
103
false
synth:7346766932921478739@1.54465
synth
[{"aliases": ["src/stemworks/rules.py", "rules.py"], "kind": "file", "value": "src/stemworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemworks/tables.py", "tables.py"], "kind": "file", "value": "src/stemworks/tables.py"}, {"ali...
[{"answers": ["src/stemworks/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-s1-00022-twin
stemworks
ctx_compress-s1-00022
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_stem double counts carried amounts when the period closes on a weekend. It lives in src/stemworks/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 stemworks ...
28,785
1,439
47
103
false
synth:7346766932921478739@1.54465
synth
[{"aliases": ["src/stemworks/rules.py", "rules.py"], "kind": "file", "value": "src/stemworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemworks/tables.py", "tables.py"], "kind": "file", "value": "src/stemworks/tables.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": ["caption_2.csv", "data/caption_2.csv"]...
ctx_compress-s1-00023
gaugekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_gauge_safe lets defaults overwrite explicit overrides. It lives in src/gaugekit/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 gaugekit layout. [tool_use kind...
68,749
3,072
109
122
false
synth:5100851436875047160@1.29345
synth
[{"aliases": ["src/gaugekit/compat.py", "compat.py"], "kind": "file", "value": "src/gaugekit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases"...
[{"answers": ["src/gaugekit/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 mos...
ctx_compress-s1-00023-twin
gaugekit
ctx_compress-s1-00023
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_gauge_safe lets defaults overwrite explicit overrides. It lives in src/gaugekit/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 gaugekit layout. [tool_use kind...
39,832
1,991
71
122
false
synth:5100851436875047160@1.29345
synth
[{"aliases": ["src/gaugekit/compat.py", "compat.py"], "kind": "file", "value": "src/gaugekit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases"...
[{"answers": ["src/gaugekit/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 mos...
ctx_compress-s1-00024
shipmentforge
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_stock raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/shipmentforge/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 ship...
75,618
3,072
106
131
false
synth:4190188320554364542@2.26688
synth
[{"aliases": ["src/shipmentforge/limits.py", "limits.py"], "kind": "file", "value": "src/shipmentforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/shipmentforge/report.py", "report.py"], "kind": "file", "value": "src/shipmentfo...
[{"answers": ["src/shipmentforge/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-s1-00024-twin
shipmentforge
ctx_compress-s1-00024
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_stock raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/shipmentforge/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 ship...
44,485
2,224
63
131
false
synth:4190188320554364542@2.26688
synth
[{"aliases": ["src/shipmentforge/limits.py", "limits.py"], "kind": "file", "value": "src/shipmentforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/shipmentforge/report.py", "report.py"], "kind": "file", "value": "src/shipmentfo...
[{"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_5.csv", "reorder_5.csv"]...
ctx_compress-s1-00025
marginflow
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_heading_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/marginflow/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 ...
95,729
3,072
91
104
false
synth:7193884536087186964@1.98142
synth
[{"aliases": ["src/marginflow/limits.py", "limits.py"], "kind": "file", "value": "src/marginflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/marginflow/compat.py", "compat.py"], "kind": "file", "value": "src/marginflow/compat.p...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/marginflow/tables.py", "tables.py...
ctx_compress-s1-00025-twin
marginflow
ctx_compress-s1-00025
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_heading_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/marginflow/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 ...
51,159
2,557
49
104
false
synth:7193884536087186964@1.98142
synth
[{"aliases": ["src/marginflow/limits.py", "limits.py"], "kind": "file", "value": "src/marginflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/marginflow/compat.py", "compat.py"], "kind": "file", "value": "src/marginflow/compat.p...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_core.py", "tests/test_core.py"],...
ctx_compress-s1-00026
glyphflow
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_caption skips the first page; page 1 returns what should be page 2. It lives in src/glyphflow/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 glyphflow layout. [t...
99,295
3,072
100
115
false
synth:8909806048695160027@2.03075
synth
[{"aliases": ["src/glyphflow/io.py", "io.py"], "kind": "file", "value": "src/glyphflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/glyphflow/schema.py", "schema.py"], "kind": "file", "value": "src/glyphflow/schema.py"}, {"aliases": ["data/glyp...
[{"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/glyphflow/tables.py", "tables.py"...
ctx_compress-s1-00026-twin
glyphflow
ctx_compress-s1-00026
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_caption skips the first page; page 1 returns what should be page 2. It lives in src/glyphflow/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 glyphflow layout. [t...
66,262
3,072
66
115
false
synth:8909806048695160027@2.03075
synth
[{"aliases": ["src/glyphflow/io.py", "io.py"], "kind": "file", "value": "src/glyphflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/glyphflow/schema.py", "schema.py"], "kind": "file", "value": "src/glyphflow/schema.py"}, {"aliases": ["data/glyp...
[{"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/glyphflow/core.py"], "...
ctx_compress-s1-00027
fixworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_corridor_all raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/fixworks/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 t...
103,515
3,072
108
127
false
synth:6541156969518600771@2.27405
synth
[{"aliases": ["src/fixworks/report.py", "report.py"], "kind": "file", "value": "src/fixworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/corridor_10.csv", "corridor_10.csv"], "kind": "file", "value": "data/corridor_10.csv"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00027-twin
fixworks
ctx_compress-s1-00027
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_corridor_all raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/fixworks/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 t...
68,279
3,072
68
127
false
synth:6541156969518600771@2.27405
synth
[{"aliases": ["src/fixworks/report.py", "report.py"], "kind": "file", "value": "src/fixworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/corridor_10.csv", "corridor_10.csv"], "kind": "file", "value": "data/corridor_10.csv"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["compat.py", "src/fixworks/compat.py"]...
ctx_compress-s1-00028
glyphdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_lexeme returns the upper bound for every value instead of clamping into the range. It lives in src/glyphdesk/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 gl...
37,920
1,896
68
75
false
synth:738086820421283613@1
synth
[{"aliases": ["src/glyphdesk/limits.py", "limits.py"], "kind": "file", "value": "src/glyphdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/glyphdesk/store.py", "store.py"], "kind": "file", "value": "src/glyphdesk/store.py"}, {"...
[{"answers": ["src/glyphdesk/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-s1-00028-twin
glyphdesk
ctx_compress-s1-00028
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_lexeme returns the upper bound for every value instead of clamping into the range. It lives in src/glyphdesk/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 gl...
18,631
1,000
40
75
false
synth:738086820421283613@1
synth
[{"aliases": ["src/glyphdesk/limits.py", "limits.py"], "kind": "file", "value": "src/glyphdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/glyphdesk/store.py", "store.py"], "kind": "file", "value": "src/glyphdesk/store.py"}, {"...
[{"answers": ["src/glyphdesk/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-s1-00029
traceworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket returns one item too few on every page. It lives in src/traceworks/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 traceworks layout. [tool_use kind=searc...
37,175
1,858
74
79
false
synth:326201410443488460@1
synth
[{"aliases": ["src/traceworks/audit.py", "audit.py"], "kind": "file", "value": "src/traceworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/sample_1.csv", "sample_1.csv"], "kind": "file", "value": "data/sample_1.csv"}, {"aliases": ...
[{"answers": ["src/traceworks/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-s1-00029-twin
traceworks
ctx_compress-s1-00029
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket returns one item too few on every page. It lives in src/traceworks/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 traceworks layout. [tool_use kind=searc...
19,592
1,000
44
79
false
synth:326201410443488460@1
synth
[{"aliases": ["src/traceworks/audit.py", "audit.py"], "kind": "file", "value": "src/traceworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/sample_1.csv", "sample_1.csv"], "kind": "file", "value": "data/sample_1.csv"}, {"aliases": ...
[{"answers": ["src/traceworks/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-s1-00030
policyforge
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_token_safe double counts carried amounts when the period closes on a weekend. It lives in src/policyforge/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 policyfor...
60,124
3,006
95
106
false
synth:4815495358551660373@1.33272
synth
[{"aliases": ["src/policyforge/cli.py", "cli.py"], "kind": "file", "value": "src/policyforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/policyforge/format.py", "format.py"], "kind": "file", "value": "src/policyforge/format.py"}, {"aliases...
[{"answers": ["src/policyforge/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-s1-00030-twin
policyforge
ctx_compress-s1-00030
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_token_safe double counts carried amounts when the period closes on a weekend. It lives in src/policyforge/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 policyfor...
29,314
1,465
49
106
false
synth:4815495358551660373@1.33272
synth
[{"aliases": ["src/policyforge/cli.py", "cli.py"], "kind": "file", "value": "src/policyforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/policyforge/format.py", "format.py"], "kind": "file", "value": "src/policyforge/format.py"}, {"aliases...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/policyforge/format.p...
ctx_compress-s1-00031
footnoteforge
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_page skips the first page; page 1 returns what should be page 2. It lives in src/footnoteforge/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 footno...
25,984
1,299
53
58
false
synth:7533086241579170109@1
synth
[{"aliases": ["src/footnoteforge/format.py", "format.py"], "kind": "file", "value": "src/footnoteforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/revision_1.csv", "revision_1.csv"], "kind": "file", "value": "data/revision_1.c...
[{"answers": ["src/footnoteforge/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'...
ctx_compress-s1-00031-twin
footnoteforge
ctx_compress-s1-00031
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_page skips the first page; page 1 returns what should be page 2. It lives in src/footnoteforge/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 footno...
15,885
1,000
29
58
false
synth:7533086241579170109@1
synth
[{"aliases": ["src/footnoteforge/format.py", "format.py"], "kind": "file", "value": "src/footnoteforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/revision_1.csv", "revision_1.csv"], "kind": "file", "value": "data/revision_1.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": ["io.py", "src/footnoteforge/io.py"], "...
ctx_compress-s1-00032
tokenforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_claim ignores the cap, so late attempts wait for minutes. It lives in src/tokenforge/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 tokenforge layout. [tool_u...
64,586
3,072
101
112
false
synth:1950066122520296614@2.32322
synth
[{"aliases": ["src/tokenforge/format.py", "format.py"], "kind": "file", "value": "src/tokenforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tokenforge/core.py", "core.py"], "kind": "file", "value": "src/tokenforge/core.py"}, {...
[{"answers": ["src/tokenforge/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-s1-00032-twin
tokenforge
ctx_compress-s1-00032
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] mint_claim ignores the cap, so late attempts wait for minutes. It lives in src/tokenforge/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 tokenforge layout. [tool_u...
37,124
1,856
63
112
false
synth:1950066122520296614@2.32322
synth
[{"aliases": ["src/tokenforge/format.py", "format.py"], "kind": "file", "value": "src/tokenforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tokenforge/core.py", "core.py"], "kind": "file", "value": "src/tokenforge/core.py"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00033
clauseforge
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_margin averages one value too many once the window is full. It lives in src/clauseforge/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 clauseforge layout....
28,780
1,439
49
58
false
synth:5778134716474239220@1
synth
[{"aliases": ["src/clauseforge/limits.py", "limits.py"], "kind": "file", "value": "src/clauseforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/clauseforge/schema.py", "schema.py"], "kind": "file", "value": "src/clauseforge/sche...
[{"answers": ["src/clauseforge/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 ...
ctx_compress-s1-00033-twin
clauseforge
ctx_compress-s1-00033
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_margin averages one value too many once the window is full. It lives in src/clauseforge/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 clauseforge layout....
14,771
1,000
31
58
false
synth:5778134716474239220@1
synth
[{"aliases": ["src/clauseforge/limits.py", "limits.py"], "kind": "file", "value": "src/clauseforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/clauseforge/schema.py", "schema.py"], "kind": "file", "value": "src/clauseforge/sche...
[{"answers": ["src/clauseforge/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 ...
ctx_compress-s1-00034
sampledesk
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] merge_sample lets defaults overwrite explicit overrides. It lives in src/sampledesk/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 sampledesk layout. [tool_use kind=...
117,159
3,072
128
143
false
synth:1065937477497010615@2.32938
synth
[{"aliases": ["src/sampledesk/audit.py", "audit.py"], "kind": "file", "value": "src/sampledesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sampledesk/tables.py", "tables.py"], "kind": "file", "value": "src/sampledesk/tables.py"}, {...
[{"answers": ["src/sampledesk/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-s1-00034-twin
sampledesk
ctx_compress-s1-00034
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] merge_sample lets defaults overwrite explicit overrides. It lives in src/sampledesk/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 sampledesk layout. [tool_use kind=...
61,923
3,072
76
143
false
synth:1065937477497010615@2.32938
synth
[{"aliases": ["src/sampledesk/audit.py", "audit.py"], "kind": "file", "value": "src/sampledesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/sampledesk/tables.py", "tables.py"], "kind": "file", "value": "src/sampledesk/tables.py"}, {...
[{"answers": ["src/sampledesk/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-s1-00035
clauseflow
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_revision returns the upper bound for every value instead of clamping into the range. It lives in src/clauseflow/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 c...
59,885
2,994
89
110
false
synth:7339144388418822045@1.34755
synth
[{"aliases": ["src/clauseflow/cli.py", "cli.py"], "kind": "file", "value": "src/clauseflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/section_5.csv", "section_5.csv"], "kind": "file", "value": "data/section_5.csv"}, {"aliases": ["data/ma...
[{"answers": ["src/clauseflow/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 re...
ctx_compress-s1-00035-twin
clauseflow
ctx_compress-s1-00035
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_revision returns the upper bound for every value instead of clamping into the range. It lives in src/clauseflow/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 c...
39,508
1,975
65
110
false
synth:7339144388418822045@1.34755
synth
[{"aliases": ["src/clauseflow/cli.py", "cli.py"], "kind": "file", "value": "src/clauseflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/section_5.csv", "section_5.csv"], "kind": "file", "value": "data/section_5.csv"}, {"aliases": ["data/ma...
[{"answers": ["src/clauseflow/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 re...
ctx_compress-s1-00036
sampleforge
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] downsample_gauge raises SeriesDriftError for bytes input in strict mode instead of decoding it. It lives in src/sampleforge/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 ...
43,109
2,155
83
92
false
synth:3872631031988286482@1
synth
[{"aliases": ["src/sampleforge/core.py", "core.py"], "kind": "file", "value": "src/sampleforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/bucket_1.csv", "bucket_1.csv"], "kind": "file", "value": "data/bucket_1.csv"}, {"aliases": ["sr...
[{"answers": ["src/sampleforge/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-s1-00036-twin
sampleforge
ctx_compress-s1-00036
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] downsample_gauge raises SeriesDriftError for bytes input in strict mode instead of decoding it. It lives in src/sampleforge/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 ...
27,755
1,387
51
92
false
synth:3872631031988286482@1
synth
[{"aliases": ["src/sampleforge/core.py", "core.py"], "kind": "file", "value": "src/sampleforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/bucket_1.csv", "bucket_1.csv"], "kind": "file", "value": "data/bucket_1.csv"}, {"aliases": ["sr...
[{"answers": ["src/sampleforge/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-s1-00037
tokenkit
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] mint_claim ignores the cap, so late attempts wait for minutes. It lives in src/tokenkit/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 tokenkit layout. [tool_use k...
42,676
2,133
85
104
false
synth:739626308205300626@1
synth
[{"aliases": ["src/tokenkit/compat.py", "compat.py"], "kind": "file", "value": "src/tokenkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/claim_0.csv", "claim_0.csv"], "kind": "file", "value": "data/claim_0.csv"}, {"aliases": [...
[{"answers": ["src/tokenkit/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 mos...
ctx_compress-s1-00037-twin
tokenkit
ctx_compress-s1-00037
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] mint_claim ignores the cap, so late attempts wait for minutes. It lives in src/tokenkit/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 tokenkit layout. [tool_use k...
24,042
1,202
51
104
false
synth:739626308205300626@1
synth
[{"aliases": ["src/tokenkit/compat.py", "compat.py"], "kind": "file", "value": "src/tokenkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/claim_0.csv", "claim_0.csv"], "kind": "file", "value": "data/claim_0.csv"}, {"aliases": [...
[{"answers": ["src/tokenkit/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 mos...
ctx_compress-s1-00038
gaugeworks
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] downsample_span double counts carried amounts when the period closes on a weekend. It lives in src/gaugeworks/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 gaugeworks lay...
45,049
2,252
98
105
false
synth:625070745657954309@1.13185
synth
[{"aliases": ["src/gaugeworks/io.py", "io.py"], "kind": "file", "value": "src/gaugeworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/gaugeworks/tables.py", "tables.py"], "kind": "file", "value": "src/gaugeworks/tables.py"}, {"aliases": ["src/g...
[{"answers": ["src/gaugeworks/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-s1-00038-twin
gaugeworks
ctx_compress-s1-00038
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] downsample_span double counts carried amounts when the period closes on a weekend. It lives in src/gaugeworks/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 gaugeworks lay...
35,632
1,781
62
105
false
synth:625070745657954309@1.13185
synth
[{"aliases": ["src/gaugeworks/io.py", "io.py"], "kind": "file", "value": "src/gaugeworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/gaugeworks/tables.py", "tables.py"], "kind": "file", "value": "src/gaugeworks/tables.py"}, {"aliases": ["src/g...
[{"answers": ["src/gaugeworks/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-s1-00039
shipmentdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_lot produces doubled dashes for consecutive separators. It lives in src/shipmentdesk/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 shipmentdesk layout. [to...
29,800
1,490
68
77
false
synth:7463073879316277055@1
synth
[{"aliases": ["src/shipmentdesk/rules.py", "rules.py"], "kind": "file", "value": "src/shipmentdesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/sku_1.csv", "sku_1.csv"], "kind": "file", "value": "data/sku_1.csv"}, {"aliases": ["dat...
[{"answers": ["src/shipmentdesk/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-s1-00039-twin
shipmentdesk
ctx_compress-s1-00039
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 produces doubled dashes for consecutive separators. It lives in src/shipmentdesk/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 shipmentdesk layout. [to...
16,575
1,000
40
77
false
synth:7463073879316277055@1
synth
[{"aliases": ["src/shipmentdesk/rules.py", "rules.py"], "kind": "file", "value": "src/shipmentdesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/sku_1.csv", "sku_1.csv"], "kind": "file", "value": "data/sku_1.csv"}, {"aliases": ["dat...
[{"answers": ["src/shipmentdesk/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-s1-00040
claimkit
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_session skips the first page; page 1 returns what should be page 2. It lives in src/claimkit/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 claimkit layout....
103,804
3,072
108
119
false
synth:6763692647076281011@2.19758
synth
[{"aliases": ["src/claimkit/limits.py", "limits.py"], "kind": "file", "value": "src/claimkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/claimkit/rules.py", "rules.py"], "kind": "file", "value": "src/claimkit/rules.py"}, {"alia...
[{"answers": ["src/claimkit/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 mos...
ctx_compress-s1-00040-twin
claimkit
ctx_compress-s1-00040
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_session skips the first page; page 1 returns what should be page 2. It lives in src/claimkit/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 claimkit layout....
49,764
2,488
54
119
false
synth:6763692647076281011@2.19758
synth
[{"aliases": ["src/claimkit/limits.py", "limits.py"], "kind": "file", "value": "src/claimkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/claimkit/rules.py", "rules.py"], "kind": "file", "value": "src/claimkit/rules.py"}, {"alia...
[{"answers": ["src/claimkit/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 mos...
ctx_compress-s1-00041
captionledger
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_caption drops override values of 0, False and empty string. It lives in src/captionledger/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 captionledger la...
49,100
2,455
90
97
false
synth:1054464263619304986@1.43945
synth
[{"aliases": ["src/captionledger/report.py", "report.py"], "kind": "file", "value": "src/captionledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/captionledger/core.py", "core.py"], "kind": "file", "value": "src/captionledger/...
[{"answers": ["src/captionledger/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'...
ctx_compress-s1-00041-twin
captionledger
ctx_compress-s1-00041
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_caption drops override values of 0, False and empty string. It lives in src/captionledger/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 captionledger la...
29,726
1,486
45
97
false
synth:1054464263619304986@1.43945
synth
[{"aliases": ["src/captionledger/report.py", "report.py"], "kind": "file", "value": "src/captionledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/captionledger/core.py", "core.py"], "kind": "file", "value": "src/captionledger/...
[{"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_core.py", "tests/test_core.py"],...
ctx_compress-s1-00042
scopeworks
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_grant_page double counts carried amounts when the period closes on a weekend. It lives in src/scopeworks/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 scopew...
83,939
3,072
116
137
false
synth:2442246746773313149@1.66132
synth
[{"aliases": ["src/scopeworks/store.py", "store.py"], "kind": "file", "value": "src/scopeworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/scope_10.csv", "scope_10.csv"], "kind": "file", "value": "data/scope_10.csv"}, {"aliases": ...
[{"answers": ["src/scopeworks/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-s1-00042-twin
scopeworks
ctx_compress-s1-00042
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_grant_page double counts carried amounts when the period closes on a weekend. It lives in src/scopeworks/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 scopew...
50,192
2,509
82
137
false
synth:2442246746773313149@1.66132
synth
[{"aliases": ["src/scopeworks/store.py", "store.py"], "kind": "file", "value": "src/scopeworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/scope_10.csv", "scope_10.csv"], "kind": "file", "value": "data/scope_10.csv"}, {"aliases": ...
[{"answers": ["src/scopeworks/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-s1-00043
stemworks
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_chunk returns one item too few on every page. It lives in src/stemworks/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 stemworks layout. [tool_use kind=sea...
73,094
3,072
116
125
false
synth:1703232847179681125@1.30032
synth
[{"aliases": ["src/stemworks/rules.py", "rules.py"], "kind": "file", "value": "src/stemworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemworks/io.py", "io.py"], "kind": "file", "value": "src/stemworks/io.py"}, {"aliases": ["dat...
[{"answers": ["src/stemworks/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-s1-00043-twin
stemworks
ctx_compress-s1-00043
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_chunk returns one item too few on every page. It lives in src/stemworks/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 stemworks layout. [tool_use kind=sea...
39,489
1,974
70
125
false
synth:1703232847179681125@1.30032
synth
[{"aliases": ["src/stemworks/rules.py", "rules.py"], "kind": "file", "value": "src/stemworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemworks/io.py", "io.py"], "kind": "file", "value": "src/stemworks/io.py"}, {"aliases": ["dat...
[{"answers": ["src/stemworks/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-s1-00044
chunkforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_chunk double counts carried amounts when the period closes on a weekend. It lives in src/chunkforge/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 chunkforge la...
115,335
3,072
129
138
false
synth:8314623675212119688@2.14228
synth
[{"aliases": ["src/chunkforge/store.py", "store.py"], "kind": "file", "value": "src/chunkforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/caption_3.csv", "caption_3.csv"], "kind": "file", "value": "data/caption_3.csv"}, {"aliases...
[{"answers": ["src/chunkforge/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-s1-00044-twin
chunkforge
ctx_compress-s1-00044
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_chunk double counts carried amounts when the period closes on a weekend. It lives in src/chunkforge/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 chunkforge la...
62,070
3,072
67
138
false
synth:8314623675212119688@2.14228
synth
[{"aliases": ["src/chunkforge/store.py", "store.py"], "kind": "file", "value": "src/chunkforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/caption_3.csv", "caption_3.csv"], "kind": "file", "value": "data/caption_3.csv"}, {"aliases...
[{"answers": ["src/chunkforge/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-s1-00045
bookingledger
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_quota double counts carried amounts when the period closes on a weekend. It lives in src/bookingledger/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 booking...
16,921
1,000
43
52
false
synth:4520894805321860014@1
synth
[{"aliases": ["src/bookingledger/store.py", "store.py"], "kind": "file", "value": "src/bookingledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/bookingledger/limits.py", "limits.py"], "kind": "file", "value": "src/bookingledger/li...
[{"answers": ["src/bookingledger/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 ...
ctx_compress-s1-00045-twin
bookingledger
ctx_compress-s1-00045
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_quota double counts carried amounts when the period closes on a weekend. It lives in src/bookingledger/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 booking...
9,211
1,000
27
52
false
synth:4520894805321860014@1
synth
[{"aliases": ["src/bookingledger/store.py", "store.py"], "kind": "file", "value": "src/bookingledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/bookingledger/limits.py", "limits.py"], "kind": "file", "value": "src/bookingledger/li...
[{"answers": ["src/bookingledger/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 ...
ctx_compress-s1-00046
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] renumber_margin leaves a trailing dash when the title ends in punctuation. It lives in src/footnotekit/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 footnotekit lay...
62,546
3,072
97
122
false
synth:5970523778210222498@2.0005
synth
[{"aliases": ["src/footnotekit/audit.py", "audit.py"], "kind": "file", "value": "src/footnotekit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/footnotekit/schema.py", "schema.py"], "kind": "file", "value": "src/footnotekit/schema.py"...
[{"answers": ["src/footnotekit/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 mo...
ctx_compress-s1-00046-twin
footnotekit
ctx_compress-s1-00046
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_margin leaves a trailing dash when the title ends in punctuation. It lives in src/footnotekit/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 footnotekit lay...
43,187
2,159
61
122
false
synth:5970523778210222498@2.0005
synth
[{"aliases": ["src/footnotekit/audit.py", "audit.py"], "kind": "file", "value": "src/footnotekit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/footnotekit/schema.py", "schema.py"], "kind": "file", "value": "src/footnotekit/schema.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": ["clause_5.csv", "data/clause_5.csv"], ...
ctx_compress-s1-00047
probekit
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] downsample_series raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/probekit/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 probe...
39,274
1,963
76
85
false
synth:8220249266059180672@1
synth
[{"aliases": ["src/probekit/core.py", "core.py"], "kind": "file", "value": "src/probekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/probekit/schema.py", "schema.py"], "kind": "file", "value": "src/probekit/schema.py"}, {"aliases": ["s...
[{"answers": ["src/probekit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-00047-twin
probekit
ctx_compress-s1-00047
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] downsample_series raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/probekit/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 probe...
27,081
1,354
48
85
false
synth:8220249266059180672@1
synth
[{"aliases": ["src/probekit/core.py", "core.py"], "kind": "file", "value": "src/probekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/probekit/schema.py", "schema.py"], "kind": "file", "value": "src/probekit/schema.py"}, {"aliases": ["s...
[{"answers": ["src/probekit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-00048
rosterledger
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_shift_page averages one value too many once the window is full. It lives in src/rosterledger/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 rosterledger layou...
76,143
3,072
102
121
false
synth:721363562274746272@1.50133
synth
[{"aliases": ["src/rosterledger/rules.py", "rules.py"], "kind": "file", "value": "src/rosterledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/slot_9.csv", "slot_9.csv"], "kind": "file", "value": "data/slot_9.csv"}, {"aliases": ["...
[{"answers": ["src/rosterledger/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-s1-00048-twin
rosterledger
ctx_compress-s1-00048
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_shift_page averages one value too many once the window is full. It lives in src/rosterledger/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 rosterledger layou...
47,245
2,362
68
121
false
synth:721363562274746272@1.50133
synth
[{"aliases": ["src/rosterledger/rules.py", "rules.py"], "kind": "file", "value": "src/rosterledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/slot_9.csv", "slot_9.csv"], "kind": "file", "value": "data/slot_9.csv"}, {"aliases": ["...
[{"answers": ["src/rosterledger/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-s1-00049
corridorkit
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] project_bearing_batch keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/corridorkit/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 corr...
50,338
2,516
73
82
false
synth:1362764802834090157@1
synth
[{"aliases": ["src/corridorkit/store.py", "store.py"], "kind": "file", "value": "src/corridorkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/corridorkit/limits.py", "limits.py"], "kind": "file", "value": "src/corridorkit/limits.py"...
[{"answers": ["src/corridorkit/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-s1-00049-twin
corridorkit
ctx_compress-s1-00049
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] project_bearing_batch keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/corridorkit/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 corr...
24,864
1,243
41
82
false
synth:1362764802834090157@1
synth
[{"aliases": ["src/corridorkit/store.py", "store.py"], "kind": "file", "value": "src/corridorkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/corridorkit/limits.py", "limits.py"], "kind": "file", "value": "src/corridorkit/limits.py"...
[{"answers": ["src/corridorkit/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...