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-s1-01658-twin
bearingworks
ctx_compress-s1-01658
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_waypoint raises LegStateError for bytes input in strict mode instead of decoding it. It lives in src/bearingworks/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 ma...
29,318
1,465
50
87
false
synth:119000594857403598@1
synth
[{"aliases": ["src/bearingworks/limits.py", "limits.py"], "kind": "file", "value": "src/bearingworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/bearingworks/store.py", "store.py"], "kind": "file", "value": "src/bearingworks/st...
[{"answers": ["src/bearingworks/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-s1-01659
gaugedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] smooth_bucket raises SampleRangeError for bytes input in strict mode instead of decoding it. It lives in src/gaugedesk/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping t...
45,628
2,281
71
82
false
synth:1859040969763403449@1
synth
[{"aliases": ["src/gaugedesk/report.py", "report.py"], "kind": "file", "value": "src/gaugedesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/gaugedesk/rules.py", "rules.py"], "kind": "file", "value": "src/gaugedesk/rules.py"}, {"...
[{"answers": ["src/gaugedesk/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-s1-01659-twin
gaugedesk
ctx_compress-s1-01659
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_bucket raises SampleRangeError for bytes input in strict mode instead of decoding it. It lives in src/gaugedesk/report.py; `python -m pytest tests/test_report.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping t...
20,495
1,024
41
82
false
synth:1859040969763403449@1
synth
[{"aliases": ["src/gaugedesk/report.py", "report.py"], "kind": "file", "value": "src/gaugedesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/gaugedesk/rules.py", "rules.py"], "kind": "file", "value": "src/gaugedesk/rules.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": ["report.py", "src/gaugedesk/report.py"...
ctx_compress-s1-01660
chunkworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_glyph ignores the cap, so late attempts wait for minutes. It lives in src/chunkworks/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 chunkworks layout. [tool_u...
73,383
3,072
88
99
false
synth:5820683336117165021@1.16335
synth
[{"aliases": ["src/chunkworks/core.py", "core.py"], "kind": "file", "value": "src/chunkworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/chunkworks/limits.py", "limits.py"], "kind": "file", "value": "src/chunkworks/limits.py"}, {"alias...
[{"answers": ["src/chunkworks/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-s1-01660-twin
chunkworks
ctx_compress-s1-01660
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_glyph ignores the cap, so late attempts wait for minutes. It lives in src/chunkworks/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 chunkworks layout. [tool_u...
41,805
2,090
56
99
false
synth:5820683336117165021@1.16335
synth
[{"aliases": ["src/chunkworks/core.py", "core.py"], "kind": "file", "value": "src/chunkworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/chunkworks/limits.py", "limits.py"], "kind": "file", "value": "src/chunkworks/limits.py"}, {"alias...
[{"answers": ["src/chunkworks/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-s1-01661
glyphforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_lexeme drops override values of 0, False and empty string. It lives in src/glyphforge/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 glyphforge layout. [t...
48,689
2,434
92
99
false
synth:4370416966771511763@1
synth
[{"aliases": ["src/glyphforge/limits.py", "limits.py"], "kind": "file", "value": "src/glyphforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["src/glyphforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01661-twin
glyphforge
ctx_compress-s1-01661
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_lexeme drops override values of 0, False and empty string. It lives in src/glyphforge/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 glyphforge layout. [t...
26,082
1,304
58
99
false
synth:4370416966771511763@1
synth
[{"aliases": ["src/glyphforge/limits.py", "limits.py"], "kind": "file", "value": "src/glyphforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["src/glyphforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01662
tokenworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_claim returns one item too few on every page. It lives in src/tokenworks/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 tokenworks layout. [tool_use kind=search...
56,289
2,814
87
96
false
synth:3496223656078587128@1
synth
[{"aliases": ["src/tokenworks/audit.py", "audit.py"], "kind": "file", "value": "src/tokenworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/grant_1.csv", "grant_1.csv"], "kind": "file", "value": "data/grant_1.csv"}, {"aliases": ["s...
[{"answers": ["src/tokenworks/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-s1-01662-twin
tokenworks
ctx_compress-s1-01662
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_claim returns one item too few on every page. It lives in src/tokenworks/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 tokenworks layout. [tool_use kind=search...
31,343
1,567
53
96
false
synth:3496223656078587128@1
synth
[{"aliases": ["src/tokenworks/audit.py", "audit.py"], "kind": "file", "value": "src/tokenworks/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/grant_1.csv", "grant_1.csv"], "kind": "file", "value": "data/grant_1.csv"}, {"aliases": ["s...
[{"answers": ["src/tokenworks/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-s1-01663
clauseflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] fold_margin ignores the cap, so late attempts wait for minutes. It lives in src/clauseflow/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 clauseflow layout. [tool_use ...
79,279
3,072
93
102
false
synth:7534734610267577141@1.60003
synth
[{"aliases": ["src/clauseflow/core.py", "core.py"], "kind": "file", "value": "src/clauseflow/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": ["src/...
[{"answers": ["src/clauseflow/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01663-twin
clauseflow
ctx_compress-s1-01663
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_margin ignores the cap, so late attempts wait for minutes. It lives in src/clauseflow/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 clauseflow layout. [tool_use ...
42,562
2,128
55
102
false
synth:7534734610267577141@1.60003
synth
[{"aliases": ["src/clauseflow/core.py", "core.py"], "kind": "file", "value": "src/clauseflow/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": ["src/...
[{"answers": ["src/clauseflow/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-s1-01664
captionworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_stem_safe produces doubled dashes for consecutive separators. It lives in src/captionworks/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 captionworks layou...
64,713
3,072
90
107
false
synth:3426687264408313938@1.45497
synth
[{"aliases": ["src/captionworks/compat.py", "compat.py"], "kind": "file", "value": "src/captionworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/captionworks/report.py", "report.py"], "kind": "file", "value": "src/captionworks/...
[{"answers": ["src/captionworks/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-s1-01664-twin
captionworks
ctx_compress-s1-01664
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_stem_safe produces doubled dashes for consecutive separators. It lives in src/captionworks/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 captionworks layou...
36,790
1,839
60
107
false
synth:3426687264408313938@1.45497
synth
[{"aliases": ["src/captionworks/compat.py", "compat.py"], "kind": "file", "value": "src/captionworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/captionworks/report.py", "report.py"], "kind": "file", "value": "src/captionworks/...
[{"answers": ["src/captionworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01665
bearingworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/bearingworks/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 mapp...
18,761
1,000
50
59
false
synth:3288854934647815016@1
synth
[{"aliases": ["src/bearingworks/limits.py", "limits.py"], "kind": "file", "value": "src/bearingworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"...
[{"answers": ["src/bearingworks/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01665-twin
bearingworks
ctx_compress-s1-01665
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/bearingworks/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 mapp...
12,667
1,000
32
59
false
synth:3288854934647815016@1
synth
[{"aliases": ["src/bearingworks/limits.py", "limits.py"], "kind": "file", "value": "src/bearingworks/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"...
[{"answers": ["src/bearingworks/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01666
tokenledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/tokenledger/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 tokenledge...
76,557
3,072
91
116
false
synth:3751102797905977745@1.6189
synth
[{"aliases": ["src/tokenledger/io.py", "io.py"], "kind": "file", "value": "src/tokenledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/claim_2.csv", "claim_2.csv"], "kind": "file", "value": "data/claim_2.csv"}, {"aliases": ["src/tokenledger/c...
[{"answers": ["src/tokenledger/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 rec...
ctx_compress-s1-01666-twin
tokenledger
ctx_compress-s1-01666
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/tokenledger/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 tokenledge...
37,970
1,898
55
116
false
synth:3751102797905977745@1.6189
synth
[{"aliases": ["src/tokenledger/io.py", "io.py"], "kind": "file", "value": "src/tokenledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/claim_2.csv", "claim_2.csv"], "kind": "file", "value": "data/claim_2.csv"}, {"aliases": ["src/tokenledger/c...
[{"answers": ["src/tokenledger/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01667
bucketforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_trace_batch skips the first page; page 1 returns what should be page 2. It lives in src/bucketforge/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 bucketforge layout...
70,268
3,072
106
125
false
synth:3337749428980796307@1.63285
synth
[{"aliases": ["src/bucketforge/io.py", "io.py"], "kind": "file", "value": "src/bucketforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/series_7.csv", "series_7.csv"], "kind": "file", "value": "data/series_7.csv"}, {"aliases": ["src/bucketforg...
[{"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/span_8.csv", "span_8.csv"], "kin...
ctx_compress-s1-01667-twin
bucketforge
ctx_compress-s1-01667
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_trace_batch skips the first page; page 1 returns what should be page 2. It lives in src/bucketforge/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 bucketforge layout...
44,657
2,232
70
125
false
synth:3337749428980796307@1.63285
synth
[{"aliases": ["src/bucketforge/io.py", "io.py"], "kind": "file", "value": "src/bucketforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/series_7.csv", "series_7.csv"], "kind": "file", "value": "data/series_7.csv"}, {"aliases": ["src/bucketforg...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_store.py", "tests/test_store.py"...
ctx_compress-s1-01668
bucketledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_counter_once double counts carried amounts when the period closes on a weekend. It lives in src/bucketledger/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 bu...
48,836
2,441
86
103
false
synth:7653183359438166131@1
synth
[{"aliases": ["src/bucketledger/compat.py", "compat.py"], "kind": "file", "value": "src/bucketledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/bucketledger/cli.py", "cli.py"], "kind": "file", "value": "src/bucketledger/cli.py...
[{"answers": ["src/bucketledger/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-s1-01668-twin
bucketledger
ctx_compress-s1-01668
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_counter_once double counts carried amounts when the period closes on a weekend. It lives in src/bucketledger/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 bu...
24,216
1,210
52
103
false
synth:7653183359438166131@1
synth
[{"aliases": ["src/bucketledger/compat.py", "compat.py"], "kind": "file", "value": "src/bucketledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/bucketledger/cli.py", "cli.py"], "kind": "file", "value": "src/bucketledger/cli.py...
[{"answers": ["src/bucketledger/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-s1-01669
scopekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_nonce keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/scopekit/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 scopekit layou...
56,635
2,831
88
95
false
synth:2833369192306978387@1.31815
synth
[{"aliases": ["src/scopekit/limits.py", "limits.py"], "kind": "file", "value": "src/scopekit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/scopekit/cli.py", "cli.py"], "kind": "file", "value": "src/scopekit/cli.py"}, {"aliases": ...
[{"answers": ["src/scopekit/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-01669-twin
scopekit
ctx_compress-s1-01669
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_nonce keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/scopekit/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 scopekit layou...
35,496
1,774
56
95
false
synth:2833369192306978387@1.31815
synth
[{"aliases": ["src/scopekit/limits.py", "limits.py"], "kind": "file", "value": "src/scopekit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/scopekit/cli.py", "cli.py"], "kind": "file", "value": "src/scopekit/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-s1-01670
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_tile rejects a range where lo equals hi, which is a valid single point range. It lives in src/legflow/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 legflow l...
69,865
3,072
100
117
false
synth:6799293701208380733@1.7709
synth
[{"aliases": ["src/legflow/format.py", "format.py"], "kind": "file", "value": "src/legflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/beacon_1.csv", "beacon_1.csv"], "kind": "file", "value": "data/beacon_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": ["schema.py", "src/legflow/schema.py"],...
ctx_compress-s1-01670-twin
legflow
ctx_compress-s1-01670
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_tile rejects a range where lo equals hi, which is a valid single point range. It lives in src/legflow/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 legflow l...
42,228
2,111
68
117
false
synth:6799293701208380733@1.7709
synth
[{"aliases": ["src/legflow/format.py", "format.py"], "kind": "file", "value": "src/legflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/beacon_1.csv", "beacon_1.csv"], "kind": "file", "value": "data/beacon_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": ["bearing_5.csv", "data/bearing_5.csv"]...
ctx_compress-s1-01671
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] settle_invoice produces doubled dashes for consecutive separators. It lives in src/invoicekit/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 invoicekit layout. [tool_use...
98,104
3,072
98
125
false
synth:3667665103340743267@1.95825
synth
[{"aliases": ["src/invoicekit/cli.py", "cli.py"], "kind": "file", "value": "src/invoicekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/invoicekit/schema.py", "schema.py"], "kind": "file", "value": "src/invoicekit/schema.py"}, {"aliases": [...
[{"answers": ["src/invoicekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-01671-twin
invoicekit
ctx_compress-s1-01671
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_invoice produces doubled dashes for consecutive separators. It lives in src/invoicekit/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 invoicekit layout. [tool_use...
62,243
3,072
62
125
false
synth:3667665103340743267@1.95825
synth
[{"aliases": ["src/invoicekit/cli.py", "cli.py"], "kind": "file", "value": "src/invoicekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/invoicekit/schema.py", "schema.py"], "kind": "file", "value": "src/invoicekit/schema.py"}, {"aliases": [...
[{"answers": ["src/invoicekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-01672
cadenceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_window_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/cadenceledger/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 ...
46,401
2,320
74
85
false
synth:5040861351100476573@1
synth
[{"aliases": ["src/cadenceledger/core.py", "core.py"], "kind": "file", "value": "src/cadenceledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/cadenceledger/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceledger/rules.py"}...
[{"answers": ["src/cadenceledger/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 mo...
ctx_compress-s1-01672-twin
cadenceledger
ctx_compress-s1-01672
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_window_all raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/cadenceledger/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 ...
19,242
1,000
40
85
false
synth:5040861351100476573@1
synth
[{"aliases": ["src/cadenceledger/core.py", "core.py"], "kind": "file", "value": "src/cadenceledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/cadenceledger/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceledger/rules.py"}...
[{"answers": ["src/cadenceledger/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 mo...
ctx_compress-s1-01673
quotaflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_window_strict ignores the cap, so late attempts wait for minutes. It lives in src/quotaflow/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 quotaflow layout....
78,256
3,072
95
112
false
synth:3705231373874148481@2.1076
synth
[{"aliases": ["src/quotaflow/compat.py", "compat.py"], "kind": "file", "value": "src/quotaflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/quotaflow/schema.py", "schema.py"], "kind": "file", "value": "src/quotaflow/schema.py"},...
[{"answers": ["src/quotaflow/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 mo...
ctx_compress-s1-01673-twin
quotaflow
ctx_compress-s1-01673
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window_strict ignores the cap, so late attempts wait for minutes. It lives in src/quotaflow/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 quotaflow layout....
40,783
2,039
55
112
false
synth:3705231373874148481@2.1076
synth
[{"aliases": ["src/quotaflow/compat.py", "compat.py"], "kind": "file", "value": "src/quotaflow/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/quotaflow/schema.py", "schema.py"], "kind": "file", "value": "src/quotaflow/schema.py"},...
[{"answers": ["src/quotaflow/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 mo...
ctx_compress-s1-01674
legledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_fix returns one item too few on every page. It lives in src/legledger/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 legledger layout. [tool_use kind=search n...
51,031
2,551
68
85
false
synth:2258660357131674021@1.04347
synth
[{"aliases": ["src/legledger/format.py", "format.py"], "kind": "file", "value": "src/legledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/legledger/schema.py", "schema.py"], "kind": "file", "value": "src/legledger/schema.py"},...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/legledger/core.py"], "...
ctx_compress-s1-01674-twin
legledger
ctx_compress-s1-01674
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] snap_fix returns one item too few on every page. It lives in src/legledger/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 legledger layout. [tool_use kind=search n...
34,448
1,722
46
85
false
synth:2258660357131674021@1.04347
synth
[{"aliases": ["src/legledger/format.py", "format.py"], "kind": "file", "value": "src/legledger/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/legledger/schema.py", "schema.py"], "kind": "file", "value": "src/legledger/schema.py"},...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["limits.py", "src/legledger/limits.py"...
ctx_compress-s1-01675
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] bin_span misses keys that differ only in case. 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 probeflow layout. [tool_use kind=search nam...
96,522
3,072
106
135
false
synth:2608041756919978180@1.77893
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_2.csv", "bucket_2.csv"], "kind": "file", "value": "data/bucket_2.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/sample_10.csv", "sample_10.csv"]...
ctx_compress-s1-01675-twin
probeflow
ctx_compress-s1-01675
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 misses keys that differ only in case. 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 probeflow layout. [tool_use kind=search nam...
49,652
2,482
66
135
false
synth:2608041756919978180@1.77893
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_2.csv", "bucket_2.csv"], "kind": "file", "value": "data/bucket_2.csv"}, {"aliase...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["bucket_9.csv", "data/bucket_9.csv"], ...
ctx_compress-s1-01676
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] net_credit_safe drops override values of 0, False and empty string. It lives in src/refundledger/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 refundledger layout...
21,770
1,088
33
42
false
synth:9008766357022676416@1
synth
[{"aliases": ["src/refundledger/compat.py", "compat.py"], "kind": "file", "value": "src/refundledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/refundledger/schema.py", "schema.py"], "kind": "file", "value": "src/refundledger/...
[{"answers": ["src/refundledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01676-twin
refundledger
ctx_compress-s1-01676
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_credit_safe drops override values of 0, False and empty string. It lives in src/refundledger/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 refundledger layout...
13,654
1,000
19
42
false
synth:9008766357022676416@1
synth
[{"aliases": ["src/refundledger/compat.py", "compat.py"], "kind": "file", "value": "src/refundledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/refundledger/schema.py", "schema.py"], "kind": "file", "value": "src/refundledger/...
[{"answers": ["no"], "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 test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["compat.py", "src/refundledger/compat....
ctx_compress-s1-01677
slugforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_snippet grows linearly instead of doubling. It lives in src/slugforge/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 slugforge layout. [tool_use kind=search...
35,255
1,762
58
65
false
synth:3584043596576870600@1
synth
[{"aliases": ["src/slugforge/limits.py", "limits.py"], "kind": "file", "value": "src/slugforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/cue_2.csv", "cue_2.csv"], "kind": "file", "value": "data/cue_2.csv"}, {"aliases": ["src...
[{"answers": ["src/slugforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01677-twin
slugforge
ctx_compress-s1-01677
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_snippet grows linearly instead of doubling. It lives in src/slugforge/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 slugforge layout. [tool_use kind=search...
19,353
1,000
38
65
false
synth:3584043596576870600@1
synth
[{"aliases": ["src/slugforge/limits.py", "limits.py"], "kind": "file", "value": "src/slugforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/cue_2.csv", "cue_2.csv"], "kind": "file", "value": "data/cue_2.csv"}, {"aliases": ["src...
[{"answers": ["src/slugforge/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-s1-01678
nonceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_token produces doubled dashes for consecutive separators. It lives in src/nonceledger/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 nonceledger layout. [tool...
52,600
2,630
94
101
false
synth:3752352911502597578@1
synth
[{"aliases": ["src/nonceledger/rules.py", "rules.py"], "kind": "file", "value": "src/nonceledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/nonceledger/format.py", "format.py"], "kind": "file", "value": "src/nonceledger/format.py"...
[{"answers": ["src/nonceledger/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01678-twin
nonceledger
ctx_compress-s1-01678
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_token produces doubled dashes for consecutive separators. It lives in src/nonceledger/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 nonceledger layout. [tool...
22,625
1,131
46
101
false
synth:3752352911502597578@1
synth
[{"aliases": ["src/nonceledger/rules.py", "rules.py"], "kind": "file", "value": "src/nonceledger/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/nonceledger/format.py", "format.py"], "kind": "file", "value": "src/nonceledger/format.py"...
[{"answers": ["src/nonceledger/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01679
revisionforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_index ignores the cap, so late attempts wait for minutes. It lives in src/revisionforge/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 revisionforge layout....
84,261
3,072
95
120
false
synth:693147927491105433@1.61452
synth
[{"aliases": ["src/revisionforge/format.py", "format.py"], "kind": "file", "value": "src/revisionforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/margin_1.csv", "margin_1.csv"], "kind": "file", "value": "data/margin_1.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["audit.py", "src/revisionforge/audit.p...
ctx_compress-s1-01679-twin
revisionforge
ctx_compress-s1-01679
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_index ignores the cap, so late attempts wait for minutes. It lives in src/revisionforge/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 revisionforge layout....
52,270
2,613
63
120
false
synth:693147927491105433@1.61452
synth
[{"aliases": ["src/revisionforge/format.py", "format.py"], "kind": "file", "value": "src/revisionforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/margin_1.csv", "margin_1.csv"], "kind": "file", "value": "data/margin_1.csv"}, ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_io.py", "tests/test_io.py"], "ki...
ctx_compress-s1-01680
fixdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_leg raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/fixdesk/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 fixdesk ...
20,228
1,011
39
50
false
synth:1670687796542128667@1
synth
[{"aliases": ["src/fixdesk/core.py", "core.py"], "kind": "file", "value": "src/fixdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/fixdesk/tables.py", "tables.py"], "kind": "file", "value": "src/fixdesk/tables.py"}, {"aliases": ["data/...
[{"answers": ["src/fixdesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01680-twin
fixdesk
ctx_compress-s1-01680
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_leg raises LegBoundError for bytes input in strict mode instead of decoding it. It lives in src/fixdesk/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 fixdesk ...
14,104
1,000
29
50
false
synth:1670687796542128667@1
synth
[{"aliases": ["src/fixdesk/core.py", "core.py"], "kind": "file", "value": "src/fixdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/fixdesk/tables.py", "tables.py"], "kind": "file", "value": "src/fixdesk/tables.py"}, {"aliases": ["data/...
[{"answers": ["src/fixdesk/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-s1-01681
tiledesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_beacon_window averages one value too many once the window is full. It lives in src/tiledesk/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 tiledesk layout. [too...
36,975
1,848
72
77
false
synth:3726757786698960703@1
synth
[{"aliases": ["src/tiledesk/rules.py", "rules.py"], "kind": "file", "value": "src/tiledesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/tiledesk/limits.py", "limits.py"], "kind": "file", "value": "src/tiledesk/limits.py"}, {"aliases...
[{"answers": ["src/tiledesk/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01681-twin
tiledesk
ctx_compress-s1-01681
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_beacon_window averages one value too many once the window is full. It lives in src/tiledesk/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 tiledesk layout. [too...
22,867
1,143
46
77
false
synth:3726757786698960703@1
synth
[{"aliases": ["src/tiledesk/rules.py", "rules.py"], "kind": "file", "value": "src/tiledesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/tiledesk/limits.py", "limits.py"], "kind": "file", "value": "src/tiledesk/limits.py"}, {"aliases...
[{"answers": ["src/tiledesk/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01682
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_caption keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/glyphkit/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 glyphkit la...
93,549
3,072
120
145
false
synth:1942519043554707889@1.69707
synth
[{"aliases": ["src/glyphkit/store.py", "store.py"], "kind": "file", "value": "src/glyphkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/lexeme_9.csv", "lexeme_9.csv"], "kind": "file", "value": "data/lexeme_9.csv"}, {"aliases": ["da...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["caption_13.csv", "data/caption_13.csv...
ctx_compress-s1-01682-twin
glyphkit
ctx_compress-s1-01682
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_caption keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/glyphkit/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 glyphkit la...
64,154
3,072
76
145
false
synth:1942519043554707889@1.69707
synth
[{"aliases": ["src/glyphkit/store.py", "store.py"], "kind": "file", "value": "src/glyphkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/lexeme_9.csv", "lexeme_9.csv"], "kind": "file", "value": "data/lexeme_9.csv"}, {"aliases": ["da...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01683
beaconworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_corridor_window returns the upper bound for every value instead of clamping into the range. It lives in src/beaconworks/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 mappin...
98,747
3,072
115
124
false
synth:1258627526464681250@1.6088
synth
[{"aliases": ["src/beaconworks/cli.py", "cli.py"], "kind": "file", "value": "src/beaconworks/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/corridor_11.csv", "corridor_11.csv"], "kind": "file", "value": "data/corridor_11.csv"}, {"aliases": [...
[{"answers": ["src/beaconworks/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01683-twin
beaconworks
ctx_compress-s1-01683
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_corridor_window returns the upper bound for every value instead of clamping into the range. It lives in src/beaconworks/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 mappin...
51,738
2,586
71
124
false
synth:1258627526464681250@1.6088
synth
[{"aliases": ["src/beaconworks/cli.py", "cli.py"], "kind": "file", "value": "src/beaconworks/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/corridor_11.csv", "corridor_11.csv"], "kind": "file", "value": "data/corridor_11.csv"}, {"aliases": [...
[{"answers": ["src/beaconworks/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-s1-01684
gaugedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] merge_span misses keys that differ only in case. It lives in src/gaugedesk/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 gaugedesk layout. [tool_use kind=search nam...
26,183
1,309
47
60
false
synth:6785253965852224957@1
synth
[{"aliases": ["src/gaugedesk/rules.py", "rules.py"], "kind": "file", "value": "src/gaugedesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/gaugedesk/core.py", "core.py"], "kind": "file", "value": "src/gaugedesk/core.py"}, {"aliases":...
[{"answers": ["src/gaugedesk/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01684-twin
gaugedesk
ctx_compress-s1-01684
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_span misses keys that differ only in case. It lives in src/gaugedesk/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 gaugedesk layout. [tool_use kind=search nam...
14,282
1,000
31
60
false
synth:6785253965852224957@1
synth
[{"aliases": ["src/gaugedesk/rules.py", "rules.py"], "kind": "file", "value": "src/gaugedesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/gaugedesk/core.py", "core.py"], "kind": "file", "value": "src/gaugedesk/core.py"}, {"aliases":...
[{"answers": ["src/gaugedesk/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-01685
sectionledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_index double counts carried amounts when the period closes on a weekend. It lives in src/sectionledger/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 sectionledger...
61,196
3,059
100
113
false
synth:4245948504086550664@2.12895
synth
[{"aliases": ["src/sectionledger/io.py", "io.py"], "kind": "file", "value": "src/sectionledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/sectionledger/cli.py", "cli.py"], "kind": "file", "value": "src/sectionledger/cli.py"}, {"aliases": ["da...
[{"answers": ["src/sectionledger/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01685-twin
sectionledger
ctx_compress-s1-01685
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_index double counts carried amounts when the period closes on a weekend. It lives in src/sectionledger/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 sectionledger...
40,907
2,045
64
113
false
synth:4245948504086550664@2.12895
synth
[{"aliases": ["src/sectionledger/io.py", "io.py"], "kind": "file", "value": "src/sectionledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/sectionledger/cli.py", "cli.py"], "kind": "file", "value": "src/sectionledger/cli.py"}, {"aliases": ["da...
[{"answers": ["src/sectionledger/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 r...
ctx_compress-s1-01686
tilekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_zone_batch rejects a range where lo equals hi, which is a valid single point range. It lives in src/tilekit/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 ...
75,028
3,072
80
89
false
synth:7013144209281956730@1.54188
synth
[{"aliases": ["src/tilekit/compat.py", "compat.py"], "kind": "file", "value": "src/tilekit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tilekit/format.py", "format.py"], "kind": "file", "value": "src/tilekit/format.py"}, {"alias...
[{"answers": ["src/tilekit/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01686-twin
tilekit
ctx_compress-s1-01686
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_zone_batch rejects a range where lo equals hi, which is a valid single point range. It lives in src/tilekit/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 ...
36,017
1,800
48
89
false
synth:7013144209281956730@1.54188
synth
[{"aliases": ["src/tilekit/compat.py", "compat.py"], "kind": "file", "value": "src/tilekit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tilekit/format.py", "format.py"], "kind": "file", "value": "src/tilekit/format.py"}, {"alias...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/zone_5.csv", "zone_5.csv"], "kin...
ctx_compress-s1-01687
cadenceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_window averages one value too many once the window is full. It lives in src/cadenceledger/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 cadenceledger layout. [tool...
65,047
3,072
95
110
false
synth:5892719886610921232@1.50205
synth
[{"aliases": ["src/cadenceledger/io.py", "io.py"], "kind": "file", "value": "src/cadenceledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/cadenceledger/core.py", "core.py"], "kind": "file", "value": "src/cadenceledger/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": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01687-twin
cadenceledger
ctx_compress-s1-01687
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] assign_window averages one value too many once the window is full. It lives in src/cadenceledger/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 cadenceledger layout. [tool...
35,092
1,754
53
110
false
synth:5892719886610921232@1.50205
synth
[{"aliases": ["src/cadenceledger/io.py", "io.py"], "kind": "file", "value": "src/cadenceledger/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/cadenceledger/core.py", "core.py"], "kind": "file", "value": "src/cadenceledger/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": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01688
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_balance_page averages one value too many once the window is full. It lives in src/receiptledger/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 receiptledger l...
105,882
3,072
109
130
false
synth:3278649934996784909@2.25813
synth
[{"aliases": ["src/receiptledger/audit.py", "audit.py"], "kind": "file", "value": "src/receiptledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/receiptledger/store.py", "store.py"], "kind": "file", "value": "src/receiptledger/stor...
[{"answers": ["src/receiptledger/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-s1-01688-twin
receiptledger
ctx_compress-s1-01688
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_page averages one value too many once the window is full. It lives in src/receiptledger/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 receiptledger l...
63,433
3,072
73
130
false
synth:3278649934996784909@2.25813
synth
[{"aliases": ["src/receiptledger/audit.py", "audit.py"], "kind": "file", "value": "src/receiptledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/receiptledger/store.py", "store.py"], "kind": "file", "value": "src/receiptledger/stor...
[{"answers": ["src/receiptledger/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-s1-01689
refunddesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_refund returns the upper bound for every value instead of clamping into the range. It lives in src/refunddesk/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 r...
90,246
3,072
110
119
false
synth:8797329484969554839@1.54808
synth
[{"aliases": ["src/refunddesk/audit.py", "audit.py"], "kind": "file", "value": "src/refunddesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/receipt_6.csv", "receipt_6.csv"], "kind": "file", "value": "data/receipt_6.csv"}, {"aliases...
[{"answers": ["src/refunddesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01689-twin
refunddesk
ctx_compress-s1-01689
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_refund returns the upper bound for every value instead of clamping into the range. It lives in src/refunddesk/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 r...
47,837
2,391
64
119
false
synth:8797329484969554839@1.54808
synth
[{"aliases": ["src/refunddesk/audit.py", "audit.py"], "kind": "file", "value": "src/refunddesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/receipt_6.csv", "receipt_6.csv"], "kind": "file", "value": "data/receipt_6.csv"}, {"aliases...
[{"answers": ["src/refunddesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01690
counterledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_series leaves a trailing dash when the title ends in punctuation. It lives in src/counterledger/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 counterledger layou...
62,652
3,072
105
118
false
synth:6957929601254027686@1.36193
synth
[{"aliases": ["src/counterledger/core.py", "core.py"], "kind": "file", "value": "src/counterledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/counterledger/tables.py", "tables.py"], "kind": "file", "value": "src/counterledger/tables.p...
[{"answers": ["src/counterledger/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 mo...
ctx_compress-s1-01690-twin
counterledger
ctx_compress-s1-01690
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rate_series leaves a trailing dash when the title ends in punctuation. It lives in src/counterledger/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 counterledger layou...
42,633
2,131
67
118
false
synth:6957929601254027686@1.36193
synth
[{"aliases": ["src/counterledger/core.py", "core.py"], "kind": "file", "value": "src/counterledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/counterledger/tables.py", "tables.py"], "kind": "file", "value": "src/counterledger/tables.p...
[{"answers": ["src/counterledger/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 mo...
ctx_compress-s1-01691
headingkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_draft returns the upper bound for every value instead of clamping into the range. It lives in src/headingkit/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 ...
48,977
2,448
92
101
false
synth:4572705772135636231@1.24345
synth
[{"aliases": ["src/headingkit/compat.py", "compat.py"], "kind": "file", "value": "src/headingkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/headingkit/audit.py", "audit.py"], "kind": "file", "value": "src/headingkit/audit.py"}...
[{"answers": ["src/headingkit/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01691-twin
headingkit
ctx_compress-s1-01691
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_draft returns the upper bound for every value instead of clamping into the range. It lives in src/headingkit/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 ...
24,261
1,213
55
101
false
synth:4572705772135636231@1.24345
synth
[{"aliases": ["src/headingkit/compat.py", "compat.py"], "kind": "file", "value": "src/headingkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/headingkit/audit.py", "audit.py"], "kind": "file", "value": "src/headingkit/audit.py"}...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/section_1.csv", "section_1.csv"]...
ctx_compress-s1-01692
windowflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_roster raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/windowflow/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 wi...
56,431
2,821
93
108
false
synth:708634458683651013@1.44027
synth
[{"aliases": ["src/windowflow/report.py", "report.py"], "kind": "file", "value": "src/windowflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/windowflow/store.py", "store.py"], "kind": "file", "value": "src/windowflow/store.py"}...
[{"answers": ["src/windowflow/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-s1-01692-twin
windowflow
ctx_compress-s1-01692
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_roster raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/windowflow/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 wi...
35,005
1,750
53
108
false
synth:708634458683651013@1.44027
synth
[{"aliases": ["src/windowflow/report.py", "report.py"], "kind": "file", "value": "src/windowflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/windowflow/store.py", "store.py"], "kind": "file", "value": "src/windowflow/store.py"}...
[{"answers": ["src/windowflow/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-s1-01693
binflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_lot_window leaves a trailing dash when the title ends in punctuation. It lives in src/binflow/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 binflow layout. [t...
65,057
3,072
87
106
false
synth:39767634822051669@1.97168
synth
[{"aliases": ["src/binflow/rules.py", "rules.py"], "kind": "file", "value": "src/binflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/binflow/schema.py", "schema.py"], "kind": "file", "value": "src/binflow/schema.py"}, {"aliases": [...
[{"answers": ["src/binflow/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01693-twin
binflow
ctx_compress-s1-01693
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window leaves a trailing dash when the title ends in punctuation. It lives in src/binflow/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 binflow layout. [t...
36,170
1,808
49
106
false
synth:39767634822051669@1.97168
synth
[{"aliases": ["src/binflow/rules.py", "rules.py"], "kind": "file", "value": "src/binflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/binflow/schema.py", "schema.py"], "kind": "file", "value": "src/binflow/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/stock_4.csv", "stock_4.csv"], "k...
ctx_compress-s1-01694
marginforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_draft skips the first page; page 1 returns what should be page 2. It lives in src/marginforge/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 marginforge l...
42,934
2,146
75
90
false
synth:4082371996043261244@1.25135
synth
[{"aliases": ["src/marginforge/format.py", "format.py"], "kind": "file", "value": "src/marginforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/section_1.csv", "section_1.csv"], "kind": "file", "value": "data/section_1.csv"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/footnote_0.csv", "footnote_0.csv...
ctx_compress-s1-01694-twin
marginforge
ctx_compress-s1-01694
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_draft skips the first page; page 1 returns what should be page 2. It lives in src/marginforge/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 marginforge l...
25,864
1,293
45
90
false
synth:4082371996043261244@1.25135
synth
[{"aliases": ["src/marginforge/format.py", "format.py"], "kind": "file", "value": "src/marginforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/section_1.csv", "section_1.csv"], "kind": "file", "value": "data/section_1.csv"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/section_1.csv", "section_1.csv"]...
ctx_compress-s1-01695
payoutdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_credit_strict raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/payoutdesk/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 ...
39,077
1,953
72
83
false
synth:8512640623349890267@1
synth
[{"aliases": ["src/payoutdesk/cli.py", "cli.py"], "kind": "file", "value": "src/payoutdesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/payoutdesk/io.py", "io.py"], "kind": "file", "value": "src/payoutdesk/io.py"}, {"aliases": ["src/payoutd...
[{"answers": ["src/payoutdesk/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most re...
ctx_compress-s1-01695-twin
payoutdesk
ctx_compress-s1-01695
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_credit_strict raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/payoutdesk/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 ...
19,367
1,000
39
83
false
synth:8512640623349890267@1
synth
[{"aliases": ["src/payoutdesk/cli.py", "cli.py"], "kind": "file", "value": "src/payoutdesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/payoutdesk/io.py", "io.py"], "kind": "file", "value": "src/payoutdesk/io.py"}, {"aliases": ["src/payoutd...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["cli.py", "src/payoutdesk/cli.py"], "k...
ctx_compress-s1-01696
lexemeworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/lexemeworks/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 l...
63,864
3,072
91
102
false
synth:77504017026535030@1.32695
synth
[{"aliases": ["src/lexemeworks/report.py", "report.py"], "kind": "file", "value": "src/lexemeworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_2.csv", "snippet_2.csv"], "kind": "file", "value": "data/snippet_2.csv"}, {...
[{"answers": ["src/lexemeworks/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-s1-01696-twin
lexemeworks
ctx_compress-s1-01696
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_slug raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/lexemeworks/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 l...
34,900
1,745
51
102
false
synth:77504017026535030@1.32695
synth
[{"aliases": ["src/lexemeworks/report.py", "report.py"], "kind": "file", "value": "src/lexemeworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_2.csv", "snippet_2.csv"], "kind": "file", "value": "data/snippet_2.csv"}, {...
[{"answers": ["src/lexemeworks/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-s1-01697
snippetdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] truncate_chunk lets defaults overwrite explicit overrides. It lives in src/snippetdesk/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 snippetdesk layout. [tool_use...
18,866
1,000
42
49
false
synth:7099123515434343206@1
synth
[{"aliases": ["src/snippetdesk/report.py", "report.py"], "kind": "file", "value": "src/snippetdesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/snippetdesk/rules.py", "rules.py"], "kind": "file", "value": "src/snippetdesk/rules....
[{"answers": ["src/snippetdesk/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-s1-01697-twin
snippetdesk
ctx_compress-s1-01697
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 lets defaults overwrite explicit overrides. It lives in src/snippetdesk/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 snippetdesk layout. [tool_use...
14,780
1,000
28
49
false
synth:7099123515434343206@1
synth
[{"aliases": ["src/snippetdesk/report.py", "report.py"], "kind": "file", "value": "src/snippetdesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/snippetdesk/rules.py", "rules.py"], "kind": "file", "value": "src/snippetdesk/rules....
[{"answers": ["src/snippetdesk/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-s1-01698
payoutflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_receipt averages one value too many once the window is full. It lives in src/payoutflow/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 payoutflow layout. [tool_u...
93,041
3,072
94
117
false
synth:7385069009121273582@2.05147
synth
[{"aliases": ["src/payoutflow/rules.py", "rules.py"], "kind": "file", "value": "src/payoutflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/payoutflow/limits.py", "limits.py"], "kind": "file", "value": "src/payoutflow/limits.py"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-01698-twin
payoutflow
ctx_compress-s1-01698
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_receipt averages one value too many once the window is full. It lives in src/payoutflow/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 payoutflow layout. [tool_u...
55,570
2,778
70
117
false
synth:7385069009121273582@2.05147
synth
[{"aliases": ["src/payoutflow/rules.py", "rules.py"], "kind": "file", "value": "src/payoutflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/payoutflow/limits.py", "limits.py"], "kind": "file", "value": "src/payoutflow/limits.py"}, {...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/payout_2.csv", "payout_2.csv"], ...
ctx_compress-s1-01699
sectionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_strict double counts carried amounts when the period closes on a weekend. It lives in src/sectionkit/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 ...
50,398
2,519
90
107
false
synth:2752058543492743722@1.33395
synth
[{"aliases": ["src/sectionkit/format.py", "format.py"], "kind": "file", "value": "src/sectionkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/sectionkit/schema.py", "schema.py"], "kind": "file", "value": "src/sectionkit/schema.p...
[{"answers": ["src/sectionkit/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 m...
ctx_compress-s1-01699-twin
sectionkit
ctx_compress-s1-01699
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_strict double counts carried amounts when the period closes on a weekend. It lives in src/sectionkit/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 ...
34,486
1,724
63
107
false
synth:2752058543492743722@1.33395
synth
[{"aliases": ["src/sectionkit/format.py", "format.py"], "kind": "file", "value": "src/sectionkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/sectionkit/schema.py", "schema.py"], "kind": "file", "value": "src/sectionkit/schema.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": ["format.py", "src/sectionkit/format.py...
ctx_compress-s1-01700
gaugeflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_gauge_batch double counts carried amounts when the period closes on a weekend. It lives in src/gaugeflow/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 gaugefl...
98,997
3,072
137
146
false
synth:445619340331431365@2.25292
synth
[{"aliases": ["src/gaugeflow/rules.py", "rules.py"], "kind": "file", "value": "src/gaugeflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/bucket_3.csv", "bucket_3.csv"], "kind": "file", "value": "data/bucket_3.csv"}, {"aliases": ["...
[{"answers": ["src/gaugeflow/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01700-twin
gaugeflow
ctx_compress-s1-01700
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_gauge_batch double counts carried amounts when the period closes on a weekend. It lives in src/gaugeflow/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 gaugefl...
48,476
2,423
73
146
false
synth:445619340331431365@2.25292
synth
[{"aliases": ["src/gaugeflow/rules.py", "rules.py"], "kind": "file", "value": "src/gaugeflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/bucket_3.csv", "bucket_3.csv"], "kind": "file", "value": "data/bucket_3.csv"}, {"aliases": ["...
[{"answers": ["src/gaugeflow/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-01701
bearingkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_tile skips the first page; page 1 returns what should be page 2. It lives in src/bearingkit/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 bearingkit layout...
86,791
3,072
109
126
false
synth:6039392388372858938@2.05103
synth
[{"aliases": ["src/bearingkit/audit.py", "audit.py"], "kind": "file", "value": "src/bearingkit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/fix_5.csv", "fix_5.csv"], "kind": "file", "value": "data/fix_5.csv"}, {"aliases": ["src/bea...
[{"answers": ["src/bearingkit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01701-twin
bearingkit
ctx_compress-s1-01701
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_tile skips the first page; page 1 returns what should be page 2. It lives in src/bearingkit/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 bearingkit layout...
57,175
2,858
75
126
false
synth:6039392388372858938@2.05103
synth
[{"aliases": ["src/bearingkit/audit.py", "audit.py"], "kind": "file", "value": "src/bearingkit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/fix_5.csv", "fix_5.csv"], "kind": "file", "value": "data/fix_5.csv"}, {"aliases": ["src/bea...
[{"answers": ["src/bearingkit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01702
tariffforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_credit returns one item too few on every page. It lives in src/tariffforge/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 tariffforge layout. [tool_use kind=se...
27,008
1,350
56
61
false
synth:3354691163699662903@1
synth
[{"aliases": ["src/tariffforge/limits.py", "limits.py"], "kind": "file", "value": "src/tariffforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/tariffforge/audit.py", "audit.py"], "kind": "file", "value": "src/tariffforge/audit....
[{"answers": ["src/tariffforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01702-twin
tariffforge
ctx_compress-s1-01702
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_credit returns one item too few on every page. It lives in src/tariffforge/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 tariffforge layout. [tool_use kind=se...
9,998
1,000
30
61
false
synth:3354691163699662903@1
synth
[{"aliases": ["src/tariffforge/limits.py", "limits.py"], "kind": "file", "value": "src/tariffforge/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/tariffforge/audit.py", "audit.py"], "kind": "file", "value": "src/tariffforge/audit....
[{"answers": ["src/tariffforge/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's ...
ctx_compress-s1-01703
draftflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_revision keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/draftflow/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 draftflow la...
91,154
3,072
110
133
false
synth:3007495273700076482@1.95185
synth
[{"aliases": ["src/draftflow/core.py", "core.py"], "kind": "file", "value": "src/draftflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/footnote_3.csv", "footnote_3.csv"], "kind": "file", "value": "data/footnote_3.csv"}, {"aliases": ["...
[{"answers": ["src/draftflow/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01703-twin
draftflow
ctx_compress-s1-01703
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_revision keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/draftflow/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 draftflow la...
52,305
2,615
68
133
false
synth:3007495273700076482@1.95185
synth
[{"aliases": ["src/draftflow/core.py", "core.py"], "kind": "file", "value": "src/draftflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/footnote_3.csv", "footnote_3.csv"], "kind": "file", "value": "data/footnote_3.csv"}, {"aliases": ["...
[{"answers": ["src/draftflow/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most r...
ctx_compress-s1-01704
zoneflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] project_zone double counts carried amounts when the period closes on a weekend. It lives in src/zoneflow/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the zoneflow layo...
44,397
2,219
72
89
false
synth:6644456188736243680@1
synth
[{"aliases": ["src/zoneflow/rules.py", "rules.py"], "kind": "file", "value": "src/zoneflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/zoneflow/audit.py", "audit.py"], "kind": "file", "value": "src/zoneflow/audit.py"}, {"aliases": ...
[{"answers": ["src/zoneflow/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01704-twin
zoneflow
ctx_compress-s1-01704
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_zone double counts carried amounts when the period closes on a weekend. It lives in src/zoneflow/rules.py; `python -m pytest tests/test_rules.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the zoneflow layo...
24,904
1,245
52
89
false
synth:6644456188736243680@1
synth
[{"aliases": ["src/zoneflow/rules.py", "rules.py"], "kind": "file", "value": "src/zoneflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/zoneflow/audit.py", "audit.py"], "kind": "file", "value": "src/zoneflow/audit.py"}, {"aliases": ...
[{"answers": ["src/zoneflow/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01705
scopeflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session_batch raises KeyError for a missing key instead of returning the default. It lives in src/scopeflow/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 scopeflow l...
68,896
3,072
87
106
false
synth:6780540473370633780@1.52862
synth
[{"aliases": ["src/scopeflow/io.py", "io.py"], "kind": "file", "value": "src/scopeflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_9.csv", "grant_9.csv"], "kind": "file", "value": "data/grant_9.csv"}, {"aliases": ["src/scopeflow/audit.p...
[{"answers": ["src/scopeflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recen...
ctx_compress-s1-01705-twin
scopeflow
ctx_compress-s1-01705
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session_batch raises KeyError for a missing key instead of returning the default. It lives in src/scopeflow/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 scopeflow l...
45,736
2,286
61
106
false
synth:6780540473370633780@1.52862
synth
[{"aliases": ["src/scopeflow/io.py", "io.py"], "kind": "file", "value": "src/scopeflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/grant_9.csv", "grant_9.csv"], "kind": "file", "value": "data/grant_9.csv"}, {"aliases": ["src/scopeflow/audit.p...
[{"answers": ["src/scopeflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recen...
ctx_compress-s1-01706
sectionledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_section lets defaults overwrite explicit overrides. It lives in src/sectionledger/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 sectionledger layout. [too...
80,232
3,072
98
109
false
synth:307701192316171303@1.40155
synth
[{"aliases": ["src/sectionledger/report.py", "report.py"], "kind": "file", "value": "src/sectionledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/sectionledger/core.py", "core.py"], "kind": "file", "value": "src/sectionledger/...
[{"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/sectionledger/report...
ctx_compress-s1-01706-twin
sectionledger
ctx_compress-s1-01706
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_section lets defaults overwrite explicit overrides. It lives in src/sectionledger/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 sectionledger layout. [too...
43,392
2,169
58
109
false
synth:307701192316171303@1.40155
synth
[{"aliases": ["src/sectionledger/report.py", "report.py"], "kind": "file", "value": "src/sectionledger/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/sectionledger/core.py", "core.py"], "kind": "file", "value": "src/sectionledger/...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["clause_3.csv", "data/clause_3.csv"], ...
ctx_compress-s1-01707
sessionflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_token raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/sessionflow/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 sessionflow...
107,253
3,072
128
151
false
synth:3147513058490174488@2.0897
synth
[{"aliases": ["src/sessionflow/cli.py", "cli.py"], "kind": "file", "value": "src/sessionflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/grant_0.csv", "grant_0.csv"], "kind": "file", "value": "data/grant_0.csv"}, {"aliases": ["src/session...
[{"answers": ["src/sessionflow/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01707-twin
sessionflow
ctx_compress-s1-01707
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_token raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/sessionflow/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 sessionflow...
69,550
3,072
90
151
false
synth:3147513058490174488@2.0897
synth
[{"aliases": ["src/sessionflow/cli.py", "cli.py"], "kind": "file", "value": "src/sessionflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/grant_0.csv", "grant_0.csv"], "kind": "file", "value": "data/grant_0.csv"}, {"aliases": ["src/session...
[{"answers": ["src/sessionflow/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01708
sectionflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_revision_once returns one item too few on every page. It lives in src/sectionflow/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 sectionflow layout. [tool_use kin...
27,761
1,388
39
48
false
synth:6392307346737779222@1
synth
[{"aliases": ["src/sectionflow/io.py", "io.py"], "kind": "file", "value": "src/sectionflow/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/sectionflow/tables.py", "tables.py"], "kind": "file", "value": "src/sectionflow/tables.py"}, {"aliases": ["d...
[{"answers": ["src/sectionflow/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...