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-s1-01406-twin
ledgerdesk
ctx_compress-s1-01406
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_credit double counts carried amounts when the period closes on a weekend. It lives in src/ledgerdesk/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 ledgerdesk...
19,746
1,000
32
67
false
synth:2690244723806238319@1
synth
[{"aliases": ["src/ledgerdesk/limits.py", "limits.py"], "kind": "file", "value": "src/ledgerdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/ledgerdesk/tables.py", "tables.py"], "kind": "file", "value": "src/ledgerdesk/tables.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": ["data/tariff_2.csv", "tariff_2.csv"], ...
ctx_compress-s1-01407
stemflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 raises ValueError when a value itself contains an equals sign. It lives in src/stemflow/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 stemflow layou...
67,371
3,072
92
113
false
synth:4235670175121594311@2.14843
synth
[{"aliases": ["src/stemflow/store.py", "store.py"], "kind": "file", "value": "src/stemflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/stemflow/report.py", "report.py"], "kind": "file", "value": "src/stemflow/report.py"}, {"aliases...
[{"answers": ["src/stemflow/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 most ...
ctx_compress-s1-01407-twin
stemflow
ctx_compress-s1-01407
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 raises ValueError when a value itself contains an equals sign. It lives in src/stemflow/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 stemflow layou...
37,358
1,867
51
113
false
synth:4235670175121594311@2.14843
synth
[{"aliases": ["src/stemflow/store.py", "store.py"], "kind": "file", "value": "src/stemflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/stemflow/report.py", "report.py"], "kind": "file", "value": "src/stemflow/report.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": ["report.py", "src/stemflow/report.py"]...
ctx_compress-s1-01408
invoicekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_payout ignores the cap, so late attempts wait for minutes. It lives in src/invoicekit/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 invoicekit layout. [too...
34,025
1,701
45
56
false
synth:6103198546034098175@1
synth
[{"aliases": ["src/invoicekit/store.py", "store.py"], "kind": "file", "value": "src/invoicekit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/invoicekit/report.py", "report.py"], "kind": "file", "value": "src/invoicekit/report.py"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/invoicekit/store.py", "store.py"]...
ctx_compress-s1-01408-twin
invoicekit
ctx_compress-s1-01408
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_payout ignores the cap, so late attempts wait for minutes. It lives in src/invoicekit/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 invoicekit layout. [too...
21,845
1,092
31
56
false
synth:6103198546034098175@1
synth
[{"aliases": ["src/invoicekit/store.py", "store.py"], "kind": "file", "value": "src/invoicekit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/invoicekit/report.py", "report.py"], "kind": "file", "value": "src/invoicekit/report.py"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_audit.py", "tests/test_audit.py"...
ctx_compress-s1-01409
bucketdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_probe leaves a trailing dash when the title ends in punctuation. It lives in src/bucketdesk/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 bucketdesk layout. [t...
27,436
1,371
58
71
false
synth:2875626723334212938@1
synth
[{"aliases": ["src/bucketdesk/store.py", "store.py"], "kind": "file", "value": "src/bucketdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/bucketdesk/io.py", "io.py"], "kind": "file", "value": "src/bucketdesk/io.py"}, {"aliases": [...
[{"answers": ["src/bucketdesk/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-s1-01409-twin
bucketdesk
ctx_compress-s1-01409
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_probe leaves a trailing dash when the title ends in punctuation. It lives in src/bucketdesk/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 bucketdesk layout. [t...
13,788
1,000
36
71
false
synth:2875626723334212938@1
synth
[{"aliases": ["src/bucketdesk/store.py", "store.py"], "kind": "file", "value": "src/bucketdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/bucketdesk/io.py", "io.py"], "kind": "file", "value": "src/bucketdesk/io.py"}, {"aliases": [...
[{"answers": ["src/bucketdesk/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-s1-01410
fixledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_bearing raises ValueError when a value itself contains an equals sign. It lives in src/fixledger/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 fixledger lay...
19,297
1,000
59
64
false
synth:2157339551467801485@1
synth
[{"aliases": ["src/fixledger/store.py", "store.py"], "kind": "file", "value": "src/fixledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/fixledger/core.py", "core.py"], "kind": "file", "value": "src/fixledger/core.py"}, {"aliases":...
[{"answers": ["src/fixledger/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-s1-01410-twin
fixledger
ctx_compress-s1-01410
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_bearing raises ValueError when a value itself contains an equals sign. It lives in src/fixledger/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 fixledger lay...
8,106
1,000
31
64
false
synth:2157339551467801485@1
synth
[{"aliases": ["src/fixledger/store.py", "store.py"], "kind": "file", "value": "src/fixledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/fixledger/core.py", "core.py"], "kind": "file", "value": "src/fixledger/core.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-s1-01411
lotflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_lot averages one value too many once the window is full. It lives in src/lotflow/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 lotflow layout. [tool_use kind=se...
64,468
3,072
100
119
false
synth:6706290316510638377@1.85435
synth
[{"aliases": ["src/lotflow/core.py", "core.py"], "kind": "file", "value": "src/lotflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/palle...
[{"answers": ["src/lotflow/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 rec...
ctx_compress-s1-01411-twin
lotflow
ctx_compress-s1-01411
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_lot averages one value too many once the window is full. It lives in src/lotflow/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 lotflow layout. [tool_use kind=se...
38,197
1,909
61
119
false
synth:6706290316510638377@1.85435
synth
[{"aliases": ["src/lotflow/core.py", "core.py"], "kind": "file", "value": "src/lotflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/palle...
[{"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-s1-01412
legflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_leg lets defaults overwrite explicit overrides. It lives in src/legflow/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 legflow layout. [tool_use kind=search name=...
24,958
1,247
48
61
false
synth:2421125004847069247@1
synth
[{"aliases": ["src/legflow/core.py", "core.py"], "kind": "file", "value": "src/legflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/beari...
[{"answers": ["src/legflow/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 rec...
ctx_compress-s1-01412-twin
legflow
ctx_compress-s1-01412
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_leg lets defaults overwrite explicit overrides. It lives in src/legflow/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 legflow layout. [tool_use kind=search name=...
16,250
1,000
36
61
false
synth:2421125004847069247@1
synth
[{"aliases": ["src/legflow/core.py", "core.py"], "kind": "file", "value": "src/legflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/beari...
[{"answers": ["src/legflow/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 rec...
ctx_compress-s1-01413
revisionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_index raises FootnoteBoundError for bytes input in strict mode instead of decoding it. It lives in src/revisionkit/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 th...
82,437
3,072
99
114
false
synth:4076532809664902366@1.72
synth
[{"aliases": ["src/revisionkit/store.py", "store.py"], "kind": "file", "value": "src/revisionkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/draft_2.csv", "draft_2.csv"], "kind": "file", "value": "data/draft_2.csv"}, {"aliases": [...
[{"answers": ["src/revisionkit/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-s1-01413-twin
revisionkit
ctx_compress-s1-01413
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_index raises FootnoteBoundError for bytes input in strict mode instead of decoding it. It lives in src/revisionkit/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 th...
53,039
2,651
65
114
false
synth:4076532809664902366@1.72
synth
[{"aliases": ["src/revisionkit/store.py", "store.py"], "kind": "file", "value": "src/revisionkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/draft_2.csv", "draft_2.csv"], "kind": "file", "value": "data/draft_2.csv"}, {"aliases": [...
[{"answers": ["src/revisionkit/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-s1-01414
sampleworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/sampleworks/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 ...
40,934
2,046
63
74
false
synth:3608717613268997869@1
synth
[{"aliases": ["src/sampleworks/store.py", "store.py"], "kind": "file", "value": "src/sampleworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases"...
[{"answers": ["src/sampleworks/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01414-twin
sampleworks
ctx_compress-s1-01414
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/sampleworks/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 ...
24,601
1,230
43
74
false
synth:3608717613268997869@1
synth
[{"aliases": ["src/sampleworks/store.py", "store.py"], "kind": "file", "value": "src/sampleworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases"...
[{"answers": ["src/sampleworks/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-s1-01415
captiondesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_chunk_all ignores the cap, so late attempts wait for minutes. It lives in src/captiondesk/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 captiondesk layou...
99,399
3,072
122
149
false
synth:3343721014090526510@2.1598
synth
[{"aliases": ["src/captiondesk/report.py", "report.py"], "kind": "file", "value": "src/captiondesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/captiondesk/core.py", "core.py"], "kind": "file", "value": "src/captiondesk/core.py"...
[{"answers": ["src/captiondesk/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-s1-01415-twin
captiondesk
ctx_compress-s1-01415
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_chunk_all ignores the cap, so late attempts wait for minutes. It lives in src/captiondesk/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 captiondesk layou...
58,022
2,901
76
149
false
synth:3343721014090526510@2.1598
synth
[{"aliases": ["src/captiondesk/report.py", "report.py"], "kind": "file", "value": "src/captiondesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/captiondesk/core.py", "core.py"], "kind": "file", "value": "src/captiondesk/core.py"...
[{"answers": ["src/captiondesk/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01416
grantforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_session averages one value too many once the window is full. It lives in src/grantforge/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the grantforge layout. [too...
58,735
2,936
75
84
false
synth:3487852044374404091@1
synth
[{"aliases": ["src/grantforge/store.py", "store.py"], "kind": "file", "value": "src/grantforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/token_1.csv", "token_1.csv"], "kind": "file", "value": "data/token_1.csv"}, {"aliases": ["s...
[{"answers": ["src/grantforge/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01416-twin
grantforge
ctx_compress-s1-01416
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session averages one value too many once the window is full. It lives in src/grantforge/store.py; `python -m pytest tests/test_store.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the grantforge layout. [too...
27,355
1,367
41
84
false
synth:3487852044374404091@1
synth
[{"aliases": ["src/grantforge/store.py", "store.py"], "kind": "file", "value": "src/grantforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/token_1.csv", "token_1.csv"], "kind": "file", "value": "data/token_1.csv"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["claim_6.csv", "data/claim_6.csv"], "k...
ctx_compress-s1-01417
policydesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_scope_once double counts carried amounts when the period closes on a weekend. It lives in src/policydesk/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 policyde...
60,909
3,045
96
119
false
synth:4325831276327087181@1.5556
synth
[{"aliases": ["src/policydesk/store.py", "store.py"], "kind": "file", "value": "src/policydesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/policydesk/cli.py", "cli.py"], "kind": "file", "value": "src/policydesk/cli.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-s1-01417-twin
policydesk
ctx_compress-s1-01417
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_scope_once double counts carried amounts when the period closes on a weekend. It lives in src/policydesk/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 policyde...
40,355
2,017
64
119
false
synth:4325831276327087181@1.5556
synth
[{"aliases": ["src/policydesk/store.py", "store.py"], "kind": "file", "value": "src/policydesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/policydesk/cli.py", "cli.py"], "kind": "file", "value": "src/policydesk/cli.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": ["data/token_4.csv", "token_4.csv"], "k...
ctx_compress-s1-01418
principalforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_policy_strict double counts carried amounts when the period closes on a weekend. It lives in src/principalforge/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 t...
89,346
3,072
118
149
false
synth:6211561165941224939@2.15665
synth
[{"aliases": ["src/principalforge/compat.py", "compat.py"], "kind": "file", "value": "src/principalforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/token_11.csv", "token_11.csv"], "kind": "file", "value": "data/token_11.csv"}...
[{"answers": ["src/principalforge/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...
ctx_compress-s1-01418-twin
principalforge
ctx_compress-s1-01418
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_policy_strict double counts carried amounts when the period closes on a weekend. It lives in src/principalforge/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 t...
54,208
2,710
68
149
false
synth:6211561165941224939@2.15665
synth
[{"aliases": ["src/principalforge/compat.py", "compat.py"], "kind": "file", "value": "src/principalforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/token_11.csv", "token_11.csv"], "kind": "file", "value": "data/token_11.csv"}...
[{"answers": ["src/principalforge/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...
ctx_compress-s1-01419
grantflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/grantflow/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 gra...
72,749
3,072
101
114
false
synth:3079241217290922944@1.51725
synth
[{"aliases": ["src/grantflow/cli.py", "cli.py"], "kind": "file", "value": "src/grantflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/grantflow/tables.py", "tables.py"], "kind": "file", "value": "src/grantflow/tables.py"}, {"aliases": ["src...
[{"answers": ["src/grantflow/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rec...
ctx_compress-s1-01419-twin
grantflow
ctx_compress-s1-01419
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/grantflow/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 gra...
38,312
1,915
61
114
false
synth:3079241217290922944@1.51725
synth
[{"aliases": ["src/grantflow/cli.py", "cli.py"], "kind": "file", "value": "src/grantflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/grantflow/tables.py", "tables.py"], "kind": "file", "value": "src/grantflow/tables.py"}, {"aliases": ["src...
[{"answers": ["src/grantflow/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01420
probeledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] smooth_probe lets defaults overwrite explicit overrides. It lives in src/probeledger/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 probeledger layout. [tool_use k...
94,747
3,072
112
125
false
synth:7677564017943769541@1.9602
synth
[{"aliases": ["src/probeledger/report.py", "report.py"], "kind": "file", "value": "src/probeledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/probeledger/format.py", "format.py"], "kind": "file", "value": "src/probeledger/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": ["counter_8.csv", "data/counter_8.csv"]...
ctx_compress-s1-01420-twin
probeledger
ctx_compress-s1-01420
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_probe lets defaults overwrite explicit overrides. It lives in src/probeledger/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 probeledger layout. [tool_use k...
58,041
2,902
70
125
false
synth:7677564017943769541@1.9602
synth
[{"aliases": ["src/probeledger/report.py", "report.py"], "kind": "file", "value": "src/probeledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/probeledger/format.py", "format.py"], "kind": "file", "value": "src/probeledger/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": ["io.py", "src/probeledger/io.py"], "ki...
ctx_compress-s1-01421
warehouseworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_pallet_all grows linearly instead of doubling. It lives in src/warehouseworks/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 warehouseworks layout. [tool_use...
26,626
1,331
65
74
false
synth:4680320884499354607@1
synth
[{"aliases": ["src/warehouseworks/audit.py", "audit.py"], "kind": "file", "value": "src/warehouseworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/warehouseworks/compat.py", "compat.py"], "kind": "file", "value": "src/warehousework...
[{"answers": ["src/warehouseworks/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-s1-01421-twin
warehouseworks
ctx_compress-s1-01421
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_pallet_all grows linearly instead of doubling. It lives in src/warehouseworks/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 warehouseworks layout. [tool_use...
14,688
1,000
35
74
false
synth:4680320884499354607@1
synth
[{"aliases": ["src/warehouseworks/audit.py", "audit.py"], "kind": "file", "value": "src/warehouseworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/warehouseworks/compat.py", "compat.py"], "kind": "file", "value": "src/warehousework...
[{"answers": ["src/warehouseworks/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-s1-01422
beaconkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_zone averages one value too many once the window is full. It lives in src/beaconkit/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 beaconkit layout. [too...
58,208
2,910
79
96
false
synth:2053694143170065303@1
synth
[{"aliases": ["src/beaconkit/store.py", "store.py"], "kind": "file", "value": "src/beaconkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/corridor_3.csv", "corridor_3.csv"], "kind": "file", "value": "data/corridor_3.csv"}, {"aliase...
[{"answers": ["src/beaconkit/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 most...
ctx_compress-s1-01422-twin
beaconkit
ctx_compress-s1-01422
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_zone averages one value too many once the window is full. It lives in src/beaconkit/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 beaconkit layout. [too...
29,278
1,463
47
96
false
synth:2053694143170065303@1
synth
[{"aliases": ["src/beaconkit/store.py", "store.py"], "kind": "file", "value": "src/beaconkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/corridor_3.csv", "corridor_3.csv"], "kind": "file", "value": "data/corridor_3.csv"}, {"aliase...
[{"answers": ["src/beaconkit/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-s1-01423
principalforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/principalforge/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 princi...
36,919
1,845
66
71
false
synth:6726949973494975949@1
synth
[{"aliases": ["src/principalforge/report.py", "report.py"], "kind": "file", "value": "src/principalforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/principalforge/format.py", "format.py"], "kind": "file", "value": "src/princip...
[{"answers": ["src/principalforge/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-s1-01423-twin
principalforge
ctx_compress-s1-01423
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/principalforge/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 princi...
14,306
1,000
33
71
false
synth:6726949973494975949@1
synth
[{"aliases": ["src/principalforge/report.py", "report.py"], "kind": "file", "value": "src/principalforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/principalforge/format.py", "format.py"], "kind": "file", "value": "src/princip...
[{"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/principalforge/store.py", "store....
ctx_compress-s1-01424
receiptledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_ledger double counts carried amounts when the period closes on a weekend. It lives in src/receiptledger/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 recei...
29,079
1,453
42
45
false
synth:8719767856482282475@1
synth
[{"aliases": ["src/receiptledger/format.py", "format.py"], "kind": "file", "value": "src/receiptledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/receiptledger/tables.py", "tables.py"], "kind": "file", "value": "src/receiptled...
[{"answers": ["tests/test_format.py", "test_format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01425
legkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] project_fix ignores the cap, so late attempts wait for minutes. It lives in src/legkit/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 legkit layout. [tool_use kind=search ...
66,348
3,072
100
113
false
synth:1185592728883384493@1.42378
synth
[{"aliases": ["src/legkit/io.py", "io.py"], "kind": "file", "value": "src/legkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/legkit/tables.py", "tables.py"], "kind": "file", "value": "src/legkit/tables.py"}, {"aliases": ["data/fix_5.csv", "fix...
[{"answers": ["src/legkit/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 t...
ctx_compress-s1-01425-twin
legkit
ctx_compress-s1-01425
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_fix ignores the cap, so late attempts wait for minutes. It lives in src/legkit/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 legkit layout. [tool_use kind=search ...
43,577
2,178
62
113
false
synth:1185592728883384493@1.42378
synth
[{"aliases": ["src/legkit/io.py", "io.py"], "kind": "file", "value": "src/legkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/legkit/tables.py", "tables.py"], "kind": "file", "value": "src/legkit/tables.py"}, {"aliases": ["data/fix_5.csv", "fix...
[{"answers": ["src/legkit/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 t...
ctx_compress-s1-01426
margindesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_clause grows linearly instead of doubling. It lives in src/margindesk/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 margindesk layout. [tool_use kind=search ...
34,645
1,732
74
83
false
synth:5869434446144537644@1
synth
[{"aliases": ["src/margindesk/store.py", "store.py"], "kind": "file", "value": "src/margindesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/margindesk/format.py", "format.py"], "kind": "file", "value": "src/margindesk/format.py"}, {...
[{"answers": ["src/margindesk/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-s1-01426-twin
margindesk
ctx_compress-s1-01426
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_clause grows linearly instead of doubling. It lives in src/margindesk/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 margindesk layout. [tool_use kind=search ...
15,513
1,000
40
83
false
synth:5869434446144537644@1
synth
[{"aliases": ["src/margindesk/store.py", "store.py"], "kind": "file", "value": "src/margindesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/margindesk/format.py", "format.py"], "kind": "file", "value": "src/margindesk/format.py"}, {...
[{"answers": ["src/margindesk/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-s1-01427
revisionworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_heading raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/revisionworks/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...
84,770
3,072
101
124
false
synth:2921935658095205116@1.67358
synth
[{"aliases": ["src/revisionworks/rules.py", "rules.py"], "kind": "file", "value": "src/revisionworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/heading_5.csv", "heading_5.csv"], "kind": "file", "value": "data/heading_5.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01427-twin
revisionworks
ctx_compress-s1-01427
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_heading raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/revisionworks/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...
55,135
2,756
63
124
false
synth:2921935658095205116@1.67358
synth
[{"aliases": ["src/revisionworks/rules.py", "rules.py"], "kind": "file", "value": "src/revisionworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/heading_5.csv", "heading_5.csv"], "kind": "file", "value": "data/heading_5.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/heading_5.csv", "heading_5.csv"]...
ctx_compress-s1-01428
snippetworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 grows linearly instead of doubling. It lives in src/snippetworks/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 snippetworks layout. [tool_use kind=sear...
75,517
3,072
110
131
false
synth:8995377358861771668@1.9255
synth
[{"aliases": ["src/snippetworks/audit.py", "audit.py"], "kind": "file", "value": "src/snippetworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/lexeme_6.csv", "lexeme_6.csv"], "kind": "file", "value": "data/lexeme_6.csv"}, {"aliase...
[{"answers": ["src/snippetworks/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-s1-01428-twin
snippetworks
ctx_compress-s1-01428
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 grows linearly instead of doubling. It lives in src/snippetworks/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 snippetworks layout. [tool_use kind=sear...
37,602
1,880
59
131
false
synth:8995377358861771668@1.9255
synth
[{"aliases": ["src/snippetworks/audit.py", "audit.py"], "kind": "file", "value": "src/snippetworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/lexeme_6.csv", "lexeme_6.csv"], "kind": "file", "value": "data/lexeme_6.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["chunk_0.csv", "data/chunk_0.csv"], "k...
ctx_compress-s1-01429
balanceflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_balance returns the upper bound for every value instead of clamping into the range. It lives in src/balanceflow/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 ...
83,193
3,072
98
111
false
synth:1833186847403545698@1.75048
synth
[{"aliases": ["src/balanceflow/audit.py", "audit.py"], "kind": "file", "value": "src/balanceflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/tariff_2.csv", "tariff_2.csv"], "kind": "file", "value": "data/tariff_2.csv"}, {"aliases"...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/tariff_1.csv", "tariff_1.csv"], ...
ctx_compress-s1-01429-twin
balanceflow
ctx_compress-s1-01429
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_balance returns the upper bound for every value instead of clamping into the range. It lives in src/balanceflow/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 ...
43,306
2,165
52
111
false
synth:1833186847403545698@1.75048
synth
[{"aliases": ["src/balanceflow/audit.py", "audit.py"], "kind": "file", "value": "src/balanceflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/tariff_2.csv", "tariff_2.csv"], "kind": "file", "value": "data/tariff_2.csv"}, {"aliases"...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/balanceflow/core.py"],...
ctx_compress-s1-01430
probeflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 raises KeyError for a missing key instead of returning the default. It lives in src/probeflow/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 probefl...
76,053
3,072
93
100
false
synth:3739532086047022021@1.94087
synth
[{"aliases": ["src/probeflow/format.py", "format.py"], "kind": "file", "value": "src/probeflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/bucket_5.csv", "bucket_5.csv"], "kind": "file", "value": "data/bucket_5.csv"}, {"aliase...
[{"answers": ["src/probeflow/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-s1-01430-twin
probeflow
ctx_compress-s1-01430
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 raises KeyError for a missing key instead of returning the default. It lives in src/probeflow/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 probefl...
46,770
2,338
55
100
false
synth:3739532086047022021@1.94087
synth
[{"aliases": ["src/probeflow/format.py", "format.py"], "kind": "file", "value": "src/probeflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/bucket_5.csv", "bucket_5.csv"], "kind": "file", "value": "data/bucket_5.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["limits.py", "src/probeflow/limits.py"...
ctx_compress-s1-01431
captionforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/captionforge/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 capt...
49,523
2,476
88
95
false
synth:5592609282386555559@1.0584
synth
[{"aliases": ["src/captionforge/core.py", "core.py"], "kind": "file", "value": "src/captionforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/captionforge/format.py", "format.py"], "kind": "file", "value": "src/captionforge/format.py"},...
[{"answers": ["src/captionforge/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01431-twin
captionforge
ctx_compress-s1-01431
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/captionforge/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 capt...
24,278
1,213
49
95
false
synth:5592609282386555559@1.0584
synth
[{"aliases": ["src/captionforge/core.py", "core.py"], "kind": "file", "value": "src/captionforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/captionforge/format.py", "format.py"], "kind": "file", "value": "src/captionforge/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": ["format.py", "src/captionforge/format....
ctx_compress-s1-01432
bucketkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_span raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/bucketkit/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 bucketkit...
74,690
3,072
104
125
false
synth:8910798645180284145@1.94663
synth
[{"aliases": ["src/bucketkit/limits.py", "limits.py"], "kind": "file", "value": "src/bucketkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/series_1.csv", "series_1.csv"], "kind": "file", "value": "data/series_1.csv"}, {"aliase...
[{"answers": ["src/bucketkit/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-01432-twin
bucketkit
ctx_compress-s1-01432
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_span raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/bucketkit/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 bucketkit...
50,407
2,520
65
125
false
synth:8910798645180284145@1.94663
synth
[{"aliases": ["src/bucketkit/limits.py", "limits.py"], "kind": "file", "value": "src/bucketkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/series_1.csv", "series_1.csv"], "kind": "file", "value": "data/series_1.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["limits.py", "src/bucketkit/limits.py"...
ctx_compress-s1-01433
probeledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] downsample_gauge_once produces doubled dashes for consecutive separators. It lives in src/probeledger/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 probeledger la...
67,210
3,072
96
117
false
synth:1661255086295156307@2.31093
synth
[{"aliases": ["src/probeledger/format.py", "format.py"], "kind": "file", "value": "src/probeledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/probeledger/rules.py", "rules.py"], "kind": "file", "value": "src/probeledger/rules....
[{"answers": ["src/probeledger/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01433-twin
probeledger
ctx_compress-s1-01433
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_once produces doubled dashes for consecutive separators. It lives in src/probeledger/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 probeledger la...
53,288
2,664
69
117
false
synth:1661255086295156307@2.31093
synth
[{"aliases": ["src/probeledger/format.py", "format.py"], "kind": "file", "value": "src/probeledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/probeledger/rules.py", "rules.py"], "kind": "file", "value": "src/probeledger/rules....
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/probeledger/format.p...
ctx_compress-s1-01434
tileflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_bearing_batch skips the first page; page 1 returns what should be page 2. It lives in src/tileflow/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 tileflow layo...
59,521
2,976
95
106
false
synth:5869399558729111698@1.05775
synth
[{"aliases": ["src/tileflow/io.py", "io.py"], "kind": "file", "value": "src/tileflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/tile_3.csv", "tile_3.csv"], "kind": "file", "value": "data/tile_3.csv"}, {"aliases": ["src/tileflow/report.py", "...
[{"answers": ["src/tileflow/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-s1-01434-twin
tileflow
ctx_compress-s1-01434
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_bearing_batch skips the first page; page 1 returns what should be page 2. It lives in src/tileflow/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 tileflow layo...
31,996
1,599
51
106
false
synth:5869399558729111698@1.05775
synth
[{"aliases": ["src/tileflow/io.py", "io.py"], "kind": "file", "value": "src/tileflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/tile_3.csv", "tile_3.csv"], "kind": "file", "value": "data/tile_3.csv"}, {"aliases": ["src/tileflow/report.py", "...
[{"answers": ["src/tileflow/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-s1-01435
cadenceflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 lets defaults overwrite explicit overrides. It lives in src/cadenceflow/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 cadenceflow layout. [tool_use k...
47,868
2,393
97
104
false
synth:8081520292164869879@1
synth
[{"aliases": ["src/cadenceflow/report.py", "report.py"], "kind": "file", "value": "src/cadenceflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/cadenceflow/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceflow/rules....
[{"answers": ["src/cadenceflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01435-twin
cadenceflow
ctx_compress-s1-01435
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 lets defaults overwrite explicit overrides. It lives in src/cadenceflow/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 cadenceflow layout. [tool_use k...
22,487
1,124
47
104
false
synth:8081520292164869879@1
synth
[{"aliases": ["src/cadenceflow/report.py", "report.py"], "kind": "file", "value": "src/cadenceflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/cadenceflow/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceflow/rules....
[{"answers": ["src/cadenceflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01436
seriesflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_counter_once misses keys that differ only in case. It lives in src/seriesflow/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 seriesflow layout. [tool_use kin...
56,602
2,830
100
109
false
synth:395444911809276864@1.76618
synth
[{"aliases": ["src/seriesflow/report.py", "report.py"], "kind": "file", "value": "src/seriesflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/seriesflow/compat.py", "compat.py"], "kind": "file", "value": "src/seriesflow/compat.p...
[{"answers": ["src/seriesflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01436-twin
seriesflow
ctx_compress-s1-01436
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_counter_once misses keys that differ only in case. It lives in src/seriesflow/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 seriesflow layout. [tool_use kin...
31,029
1,551
56
109
false
synth:395444911809276864@1.76618
synth
[{"aliases": ["src/seriesflow/report.py", "report.py"], "kind": "file", "value": "src/seriesflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/seriesflow/compat.py", "compat.py"], "kind": "file", "value": "src/seriesflow/compat.p...
[{"answers": ["src/seriesflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01437
corridorledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_zone_once skips the first page; page 1 returns what should be page 2. It lives in src/corridorledger/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 corridorled...
105,372
3,072
118
141
false
synth:8261562248307205626@2.0526
synth
[{"aliases": ["src/corridorledger/io.py", "io.py"], "kind": "file", "value": "src/corridorledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/corridorledger/format.py", "format.py"], "kind": "file", "value": "src/corridorledger/format.py"}, {"a...
[{"answers": ["src/corridorledger/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01437-twin
corridorledger
ctx_compress-s1-01437
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_zone_once skips the first page; page 1 returns what should be page 2. It lives in src/corridorledger/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 corridorled...
58,968
2,948
72
141
false
synth:8261562248307205626@2.0526
synth
[{"aliases": ["src/corridorledger/io.py", "io.py"], "kind": "file", "value": "src/corridorledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/corridorledger/format.py", "format.py"], "kind": "file", "value": "src/corridorledger/format.py"}, {"a...
[{"answers": ["src/corridorledger/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01438
calendarflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_cadence raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/calendarflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
58,586
2,929
95
112
false
synth:5596081852218902139@1.79577
synth
[{"aliases": ["src/calendarflow/report.py", "report.py"], "kind": "file", "value": "src/calendarflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["booking_5.csv", "data/booking_5.csv"]...
ctx_compress-s1-01438-twin
calendarflow
ctx_compress-s1-01438
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cadence raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/calendarflow/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the ...
33,796
1,689
55
112
false
synth:5596081852218902139@1.79577
synth
[{"aliases": ["src/calendarflow/report.py", "report.py"], "kind": "file", "value": "src/calendarflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"...
[{"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-s1-01439
invoiceworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_balance raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/invoiceworks/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 ...
104,363
3,072
124
133
false
synth:8531322651986383450@1.73182
synth
[{"aliases": ["src/invoiceworks/compat.py", "compat.py"], "kind": "file", "value": "src/invoiceworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/invoiceworks/cli.py", "cli.py"], "kind": "file", "value": "src/invoiceworks/cli.py...
[{"answers": ["src/invoiceworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01439-twin
invoiceworks
ctx_compress-s1-01439
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_balance raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/invoiceworks/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 ...
53,086
2,654
72
133
false
synth:8531322651986383450@1.73182
synth
[{"aliases": ["src/invoiceworks/compat.py", "compat.py"], "kind": "file", "value": "src/invoiceworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/invoiceworks/cli.py", "cli.py"], "kind": "file", "value": "src/invoiceworks/cli.py...
[{"answers": ["src/invoiceworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01440
warehouseworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_warehouse_batch rejects a range where lo equals hi, which is a valid single point range. It lives in src/warehouseworks/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 mappi...
104,100
3,072
132
143
false
synth:8009017214690200169@2.16863
synth
[{"aliases": ["src/warehouseworks/audit.py", "audit.py"], "kind": "file", "value": "src/warehouseworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/warehouseworks/report.py", "report.py"], "kind": "file", "value": "src/warehousework...
[{"answers": ["src/warehouseworks/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-s1-01440-twin
warehouseworks
ctx_compress-s1-01440
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_warehouse_batch rejects a range where lo equals hi, which is a valid single point range. It lives in src/warehouseworks/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 mappi...
44,595
2,229
66
143
false
synth:8009017214690200169@2.16863
synth
[{"aliases": ["src/warehouseworks/audit.py", "audit.py"], "kind": "file", "value": "src/warehouseworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/warehouseworks/report.py", "report.py"], "kind": "file", "value": "src/warehousework...
[{"answers": ["src/warehouseworks/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-s1-01441
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] verify_claim_page drops override values of 0, False and empty string. It lives in src/scopeworks/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 scopeworks layout. ...
42,486
2,124
77
82
false
synth:4256714275144446922@1
synth
[{"aliases": ["src/scopeworks/report.py", "report.py"], "kind": "file", "value": "src/scopeworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/scopeworks/compat.py", "compat.py"], "kind": "file", "value": "src/scopeworks/compat.p...
[{"answers": ["src/scopeworks/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01441-twin
scopeworks
ctx_compress-s1-01441
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_claim_page drops override values of 0, False and empty string. It lives in src/scopeworks/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 scopeworks layout. ...
25,017
1,250
49
82
false
synth:4256714275144446922@1
synth
[{"aliases": ["src/scopeworks/report.py", "report.py"], "kind": "file", "value": "src/scopeworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/scopeworks/compat.py", "compat.py"], "kind": "file", "value": "src/scopeworks/compat.p...
[{"answers": ["src/scopeworks/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01442
grantledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_claim raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/grantledger/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 gra...
97,041
3,072
115
130
false
synth:8916620147657053097@2.03945
synth
[{"aliases": ["src/grantledger/compat.py", "compat.py"], "kind": "file", "value": "src/grantledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/grantledger/schema.py", "schema.py"], "kind": "file", "value": "src/grantledger/sche...
[{"answers": ["src/grantledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01442-twin
grantledger
ctx_compress-s1-01442
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_claim raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/grantledger/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 gra...
47,631
2,381
73
130
false
synth:8916620147657053097@2.03945
synth
[{"aliases": ["src/grantledger/compat.py", "compat.py"], "kind": "file", "value": "src/grantledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/grantledger/schema.py", "schema.py"], "kind": "file", "value": "src/grantledger/sche...
[{"answers": ["src/grantledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01443
grantflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_claim_page drops override values of 0, False and empty string. It lives in src/grantflow/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 grantflow layout. [t...
84,960
3,072
106
115
false
synth:9081319339978436067@1.76553
synth
[{"aliases": ["src/grantflow/report.py", "report.py"], "kind": "file", "value": "src/grantflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/grant_1.csv", "grant_1.csv"], "kind": "file", "value": "data/grant_1.csv"}, {"aliases":...
[{"answers": ["src/grantflow/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 mo...
ctx_compress-s1-01443-twin
grantflow
ctx_compress-s1-01443
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_claim_page drops override values of 0, False and empty string. It lives in src/grantflow/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 grantflow layout. [t...
36,665
1,833
52
115
false
synth:9081319339978436067@1.76553
synth
[{"aliases": ["src/grantflow/report.py", "report.py"], "kind": "file", "value": "src/grantflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/grant_1.csv", "grant_1.csv"], "kind": "file", "value": "data/grant_1.csv"}, {"aliases":...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_limits.py", "tests/test_limits.p...
ctx_compress-s1-01444
tileflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_bearing_once skips the first page; page 1 returns what should be page 2. It lives in src/tileflow/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 tileflow lay...
71,325
3,072
94
113
false
synth:3987177259447913832@1.84252
synth
[{"aliases": ["src/tileflow/store.py", "store.py"], "kind": "file", "value": "src/tileflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/fix_6.csv", "fix_6.csv"], "kind": "file", "value": "data/fix_6.csv"}, {"aliases": ["src/tileflo...
[{"answers": ["src/tileflow/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 most ...
ctx_compress-s1-01444-twin
tileflow
ctx_compress-s1-01444
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] cluster_bearing_once skips the first page; page 1 returns what should be page 2. It lives in src/tileflow/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 tileflow lay...
42,668
2,133
58
113
false
synth:3987177259447913832@1.84252
synth
[{"aliases": ["src/tileflow/store.py", "store.py"], "kind": "file", "value": "src/tileflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/fix_6.csv", "fix_6.csv"], "kind": "file", "value": "data/fix_6.csv"}, {"aliases": ["src/tileflo...
[{"answers": ["src/tileflow/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-s1-01445
ledgerflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_payout raises RefundBoundError for bytes input in strict mode instead of decoding it. It lives in src/ledgerflow/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 th...
40,351
2,017
75
86
false
synth:5938199817951720341@1
synth
[{"aliases": ["src/ledgerflow/audit.py", "audit.py"], "kind": "file", "value": "src/ledgerflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/ledgerflow/report.py", "report.py"], "kind": "file", "value": "src/ledgerflow/report.py"}, {...
[{"answers": ["src/ledgerflow/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-s1-01445-twin
ledgerflow
ctx_compress-s1-01445
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_payout raises RefundBoundError for bytes input in strict mode instead of decoding it. It lives in src/ledgerflow/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 th...
26,096
1,304
49
86
false
synth:5938199817951720341@1
synth
[{"aliases": ["src/ledgerflow/audit.py", "audit.py"], "kind": "file", "value": "src/ledgerflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/ledgerflow/report.py", "report.py"], "kind": "file", "value": "src/ledgerflow/report.py"}, {...
[{"answers": ["src/ledgerflow/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-s1-01446
shipmentledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_reorder returns one item too few on every page. It lives in src/shipmentledger/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 shipmentledger layout. [tool...
99,615
3,072
106
133
false
synth:1650583055071349922@2.26772
synth
[{"aliases": ["src/shipmentledger/report.py", "report.py"], "kind": "file", "value": "src/shipmentledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/sku_2.csv", "sku_2.csv"], "kind": "file", "value": "data/sku_2.csv"}, {"alias...
[{"answers": ["src/shipmentledger/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-s1-01446-twin
shipmentledger
ctx_compress-s1-01446
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_reorder returns one item too few on every page. It lives in src/shipmentledger/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 shipmentledger layout. [tool...
56,083
2,804
68
133
false
synth:1650583055071349922@2.26772
synth
[{"aliases": ["src/shipmentledger/report.py", "report.py"], "kind": "file", "value": "src/shipmentledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/sku_2.csv", "sku_2.csv"], "kind": "file", "value": "data/sku_2.csv"}, {"alias...
[{"answers": ["src/shipmentledger/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-s1-01447
grantforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_principal_window raises KeyError for a missing key instead of returning the default. It lives in src/grantforge/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...
36,813
1,840
67
72
false
synth:9171437357143149633@1
synth
[{"aliases": ["src/grantforge/rules.py", "rules.py"], "kind": "file", "value": "src/grantforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/token_2.csv", "token_2.csv"], "kind": "file", "value": "data/token_2.csv"}, {"aliases": ["d...
[{"answers": ["src/grantforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01447-twin
grantforge
ctx_compress-s1-01447
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window raises KeyError for a missing key instead of returning the default. It lives in src/grantforge/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...
17,402
1,000
41
72
false
synth:9171437357143149633@1
synth
[{"aliases": ["src/grantforge/rules.py", "rules.py"], "kind": "file", "value": "src/grantforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/token_2.csv", "token_2.csv"], "kind": "file", "value": "data/token_2.csv"}, {"aliases": ["d...
[{"answers": ["src/grantforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01448
tokendesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_scope misses keys that differ only in case. It lives in src/tokendesk/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 tokendesk layout. [tool_use kind=search n...
25,898
1,294
62
71
false
synth:5379622132378194671@1
synth
[{"aliases": ["src/tokendesk/audit.py", "audit.py"], "kind": "file", "value": "src/tokendesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/tokendesk/compat.py", "compat.py"], "kind": "file", "value": "src/tokendesk/compat.py"}, {"ali...
[{"answers": ["src/tokendesk/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 most...
ctx_compress-s1-01448-twin
tokendesk
ctx_compress-s1-01448
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_scope misses keys that differ only in case. It lives in src/tokendesk/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 tokendesk layout. [tool_use kind=search n...
14,441
1,000
34
71
false
synth:5379622132378194671@1
synth
[{"aliases": ["src/tokendesk/audit.py", "audit.py"], "kind": "file", "value": "src/tokendesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/tokendesk/compat.py", "compat.py"], "kind": "file", "value": "src/tokendesk/compat.py"}, {"ali...
[{"answers": ["src/tokendesk/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 most...
ctx_compress-s1-01449
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] anchor_heading_once raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/footnotekit/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 fo...
40,907
2,045
65
72
false
synth:4725843266697589993@1
synth
[{"aliases": ["src/footnotekit/cli.py", "cli.py"], "kind": "file", "value": "src/footnotekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/footnotekit/tables.py", "tables.py"], "kind": "file", "value": "src/footnotekit/tables.py"}, {"aliases...
[{"answers": ["src/footnotekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most r...
ctx_compress-s1-01449-twin
footnotekit
ctx_compress-s1-01449
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_heading_once raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/footnotekit/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 fo...
25,201
1,260
37
72
false
synth:4725843266697589993@1
synth
[{"aliases": ["src/footnotekit/cli.py", "cli.py"], "kind": "file", "value": "src/footnotekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/footnotekit/tables.py", "tables.py"], "kind": "file", "value": "src/footnotekit/tables.py"}, {"aliases...
[{"answers": ["src/footnotekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most r...
ctx_compress-s1-01450
glyphledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_once skips the first page; page 1 returns what should be page 2. It lives in src/glyphledger/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 glyphledger layo...
54,325
2,716
78
97
false
synth:8451692025908274229@1.19548
synth
[{"aliases": ["src/glyphledger/core.py", "core.py"], "kind": "file", "value": "src/glyphledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/glyphledger/io.py", "io.py"], "kind": "file", "value": "src/glyphledger/io.py"}, {"aliases": ["s...
[{"answers": ["src/glyphledger/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-s1-01450-twin
glyphledger
ctx_compress-s1-01450
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_once skips the first page; page 1 returns what should be page 2. It lives in src/glyphledger/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 glyphledger layo...
26,024
1,301
47
97
false
synth:8451692025908274229@1.19548
synth
[{"aliases": ["src/glyphledger/core.py", "core.py"], "kind": "file", "value": "src/glyphledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/glyphledger/io.py", "io.py"], "kind": "file", "value": "src/glyphledger/io.py"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/glyphledger/core.py"],...
ctx_compress-s1-01451
calendardesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cadence misses keys that differ only in case. It lives in src/calendardesk/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 calendardesk layout. [tool_use kind=searc...
32,568
1,628
62
75
false
synth:2050153340273250155@1
synth
[{"aliases": ["src/calendardesk/io.py", "io.py"], "kind": "file", "value": "src/calendardesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/calendardesk/schema.py", "schema.py"], "kind": "file", "value": "src/calendardesk/schema.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": ["data/window_3.csv", "window_3.csv"], ...
ctx_compress-s1-01451-twin
calendardesk
ctx_compress-s1-01451
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cadence misses keys that differ only in case. It lives in src/calendardesk/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 calendardesk layout. [tool_use kind=searc...
19,398
1,000
42
75
false
synth:2050153340273250155@1
synth
[{"aliases": ["src/calendardesk/io.py", "io.py"], "kind": "file", "value": "src/calendardesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/calendardesk/schema.py", "schema.py"], "kind": "file", "value": "src/calendardesk/schema.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-s1-01452
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] normalise_lexeme skips the first page; page 1 returns what should be page 2. It lives in src/glyphkit/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 glyphkit layout. [to...
54,638
2,731
77
88
false
synth:995241679613615568@1
synth
[{"aliases": ["src/glyphkit/cli.py", "cli.py"], "kind": "file", "value": "src/glyphkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/stem_1.csv", "stem_1.csv"], "kind": "file", "value": "data/stem_1.csv"}, {"aliases": ["src/glyphkit/audit.p...
[{"answers": ["src/glyphkit/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 rece...
ctx_compress-s1-01452-twin
glyphkit
ctx_compress-s1-01452
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_lexeme skips the first page; page 1 returns what should be page 2. It lives in src/glyphkit/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 glyphkit layout. [to...
30,360
1,518
41
88
false
synth:995241679613615568@1
synth
[{"aliases": ["src/glyphkit/cli.py", "cli.py"], "kind": "file", "value": "src/glyphkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/stem_1.csv", "stem_1.csv"], "kind": "file", "value": "data/stem_1.csv"}, {"aliases": ["src/glyphkit/audit.p...
[{"answers": ["src/glyphkit/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 rece...
ctx_compress-s1-01453
claimforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope skips the first page; page 1 returns what should be page 2. It lives in src/claimforge/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 claimforge layout....
53,124
2,656
87
106
false
synth:3481029653928908055@1
synth
[{"aliases": ["src/claimforge/rules.py", "rules.py"], "kind": "file", "value": "src/claimforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/claimforge/report.py", "report.py"], "kind": "file", "value": "src/claimforge/report.py"}, {...
[{"answers": ["src/claimforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01453-twin
claimforge
ctx_compress-s1-01453
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope skips the first page; page 1 returns what should be page 2. It lives in src/claimforge/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 claimforge layout....
35,566
1,778
63
106
false
synth:3481029653928908055@1
synth
[{"aliases": ["src/claimforge/rules.py", "rules.py"], "kind": "file", "value": "src/claimforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/claimforge/report.py", "report.py"], "kind": "file", "value": "src/claimforge/report.py"}, {...
[{"answers": ["src/claimforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01454
stemdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_chunk double counts carried amounts when the period closes on a weekend. It lives in src/stemdesk/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 stemdesk...
97,553
3,072
107
124
false
synth:876775828837555663@2.03563
synth
[{"aliases": ["src/stemdesk/format.py", "format.py"], "kind": "file", "value": "src/stemdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/snippet_0.csv", "snippet_0.csv"], "kind": "file", "value": "data/snippet_0.csv"}, {"alias...
[{"answers": ["src/stemdesk/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01454-twin
stemdesk
ctx_compress-s1-01454
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_chunk double counts carried amounts when the period closes on a weekend. It lives in src/stemdesk/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 stemdesk...
44,441
2,222
69
124
false
synth:876775828837555663@2.03563
synth
[{"aliases": ["src/stemdesk/format.py", "format.py"], "kind": "file", "value": "src/stemdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/snippet_0.csv", "snippet_0.csv"], "kind": "file", "value": "data/snippet_0.csv"}, {"alias...
[{"answers": ["src/stemdesk/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 mos...
ctx_compress-s1-01455
refundledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_ledger raises ValueError when a value itself contains an equals sign. It lives in src/refundledger/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 refundledger...
102,213
3,072
116
133
false
synth:4486018023653286779@2.1243
synth
[{"aliases": ["src/refundledger/limits.py", "limits.py"], "kind": "file", "value": "src/refundledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/receipt_5.csv", "receipt_5.csv"], "kind": "file", "value": "data/receipt_5.csv"},...
[{"answers": ["src/refundledger/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-s1-01455-twin
refundledger
ctx_compress-s1-01455
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_ledger raises ValueError when a value itself contains an equals sign. It lives in src/refundledger/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 refundledger...
58,078
2,903
78
133
false
synth:4486018023653286779@2.1243
synth
[{"aliases": ["src/refundledger/limits.py", "limits.py"], "kind": "file", "value": "src/refundledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/receipt_5.csv", "receipt_5.csv"], "kind": "file", "value": "data/receipt_5.csv"},...
[{"answers": ["src/refundledger/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-s1-01456
tariffflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] settle_refund ignores the cap, so late attempts wait for minutes. It lives in src/tariffflow/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 tariffflow layout. [tool_...
57,066
2,853
105
118
false
synth:8843357705971897374@1
synth
[{"aliases": ["src/tariffflow/store.py", "store.py"], "kind": "file", "value": "src/tariffflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/tariffflow/schema.py", "schema.py"], "kind": "file", "value": "src/tariffflow/schema.py"}, {...
[{"answers": ["src/tariffflow/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-s1-01456-twin
tariffflow
ctx_compress-s1-01456
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] settle_refund ignores the cap, so late attempts wait for minutes. It lives in src/tariffflow/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 tariffflow layout. [tool_...
25,181
1,259
55
118
false
synth:8843357705971897374@1
synth
[{"aliases": ["src/tariffflow/store.py", "store.py"], "kind": "file", "value": "src/tariffflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/tariffflow/schema.py", "schema.py"], "kind": "file", "value": "src/tariffflow/schema.py"}, {...
[{"answers": ["src/tariffflow/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...