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-s1-00855
captionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_caption drops override values of 0, False and empty string. It lives in src/captionkit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the captionkit layout. [tool...
59,772
2,988
94
103
false
synth:733266607344226101@1
synth
[{"aliases": ["src/captionkit/limits.py", "limits.py"], "kind": "file", "value": "src/captionkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/glyph_1.csv", "glyph_1.csv"], "kind": "file", "value": "data/glyph_1.csv"}, {"aliases...
[{"answers": ["src/captionkit/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-s1-00855-twin
captionkit
ctx_compress-s1-00855
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_caption drops override values of 0, False and empty string. It lives in src/captionkit/limits.py; `python -m pytest tests/test_limits.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the captionkit layout. [tool...
42,165
2,108
60
103
false
synth:733266607344226101@1
synth
[{"aliases": ["src/captionkit/limits.py", "limits.py"], "kind": "file", "value": "src/captionkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/glyph_1.csv", "glyph_1.csv"], "kind": "file", "value": "data/glyph_1.csv"}, {"aliases...
[{"answers": ["src/captionkit/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-00856
indexkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_heading returns the upper bound for every value instead of clamping into the range. It lives in src/indexkit/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 inde...
54,099
2,704
83
96
false
synth:5915471747086789442@1
synth
[{"aliases": ["src/indexkit/store.py", "store.py"], "kind": "file", "value": "src/indexkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/indexkit/cli.py", "cli.py"], "kind": "file", "value": "src/indexkit/cli.py"}, {"aliases": ["data...
[{"answers": ["src/indexkit/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-00856-twin
indexkit
ctx_compress-s1-00856
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_heading returns the upper bound for every value instead of clamping into the range. It lives in src/indexkit/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 inde...
31,022
1,551
51
96
false
synth:5915471747086789442@1
synth
[{"aliases": ["src/indexkit/store.py", "store.py"], "kind": "file", "value": "src/indexkit/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/indexkit/cli.py", "cli.py"], "kind": "file", "value": "src/indexkit/cli.py"}, {"aliases": ["data...
[{"answers": ["src/indexkit/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-00857
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] wrap_stem leaves a trailing dash when the title ends in punctuation. It lives in src/chunkworks/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 chunkworks layout. [to...
85,249
3,072
103
120
false
synth:5126342199997332576@1.82475
synth
[{"aliases": ["src/chunkworks/store.py", "store.py"], "kind": "file", "value": "src/chunkworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/cue_3.csv", "cue_3.csv"], "kind": "file", "value": "data/cue_3.csv"}, {"aliases": ["src/chu...
[{"answers": ["src/chunkworks/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-00857-twin
chunkworks
ctx_compress-s1-00857
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_stem leaves a trailing dash when the title ends in punctuation. It lives in src/chunkworks/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 chunkworks layout. [to...
49,936
2,496
65
120
false
synth:5126342199997332576@1.82475
synth
[{"aliases": ["src/chunkworks/store.py", "store.py"], "kind": "file", "value": "src/chunkworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/cue_3.csv", "cue_3.csv"], "kind": "file", "value": "data/cue_3.csv"}, {"aliases": ["src/chu...
[{"answers": ["src/chunkworks/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mos...
ctx_compress-s1-00858
policyflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_scope returns the upper bound for every value instead of clamping into the range. It lives in src/policyflow/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 ...
36,055
1,802
66
75
false
synth:6500359510294829356@1
synth
[{"aliases": ["src/policyflow/format.py", "format.py"], "kind": "file", "value": "src/policyflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/policyflow/tables.py", "tables.py"], "kind": "file", "value": "src/policyflow/tables.p...
[{"answers": ["src/policyflow/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-00858-twin
policyflow
ctx_compress-s1-00858
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_scope returns the upper bound for every value instead of clamping into the range. It lives in src/policyflow/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 ...
17,371
1,000
36
75
false
synth:6500359510294829356@1
synth
[{"aliases": ["src/policyflow/format.py", "format.py"], "kind": "file", "value": "src/policyflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/policyflow/tables.py", "tables.py"], "kind": "file", "value": "src/policyflow/tables.p...
[{"answers": ["src/policyflow/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-00859
spanworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_series skips the first page; page 1 returns what should be page 2. It lives in src/spanworks/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 spanworks layout...
43,131
2,156
105
112
false
synth:7274523485069401719@1.19697
synth
[{"aliases": ["src/spanworks/report.py", "report.py"], "kind": "file", "value": "src/spanworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/spanworks/tables.py", "tables.py"], "kind": "file", "value": "src/spanworks/tables.py"},...
[{"answers": ["src/spanworks/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-00859-twin
spanworks
ctx_compress-s1-00859
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_series skips the first page; page 1 returns what should be page 2. It lives in src/spanworks/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 spanworks layout...
29,114
1,455
65
112
false
synth:7274523485069401719@1.19697
synth
[{"aliases": ["src/spanworks/report.py", "report.py"], "kind": "file", "value": "src/spanworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/spanworks/tables.py", "tables.py"], "kind": "file", "value": "src/spanworks/tables.py"},...
[{"answers": ["src/spanworks/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-s1-00860
sampledesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] merge_series ignores the cap, so late attempts wait for minutes. It lives in src/sampledesk/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 sampledesk layout. [tool_use k...
37,834
1,891
82
105
false
synth:5995493536078340300@1.46037
synth
[{"aliases": ["src/sampledesk/cli.py", "cli.py"], "kind": "file", "value": "src/sampledesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/sampledesk/report.py", "report.py"], "kind": "file", "value": "src/sampledesk/report.py"}, {"aliases": [...
[{"answers": ["src/sampledesk/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-s1-00860-twin
sampledesk
ctx_compress-s1-00860
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_series ignores the cap, so late attempts wait for minutes. It lives in src/sampledesk/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 sampledesk layout. [tool_use k...
25,066
1,253
61
105
false
synth:5995493536078340300@1.46037
synth
[{"aliases": ["src/sampledesk/cli.py", "cli.py"], "kind": "file", "value": "src/sampledesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/sampledesk/report.py", "report.py"], "kind": "file", "value": "src/sampledesk/report.py"}, {"aliases": [...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["schema.py", "src/sampledesk/schema.py...
ctx_compress-s1-00861
calendarkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cadence grows linearly instead of doubling. It lives in src/calendarkit/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 calendarkit layout. [tool_use kind=s...
63,695
3,072
91
106
false
synth:7841564686383235541@1.1172
synth
[{"aliases": ["src/calendarkit/compat.py", "compat.py"], "kind": "file", "value": "src/calendarkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/quota_3.csv", "quota_3.csv"], "kind": "file", "value": "data/quota_3.csv"}, {"alias...
[{"answers": ["src/calendarkit/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-s1-00861-twin
calendarkit
ctx_compress-s1-00861
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cadence grows linearly instead of doubling. It lives in src/calendarkit/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 calendarkit layout. [tool_use kind=s...
35,984
1,799
59
106
false
synth:7841564686383235541@1.1172
synth
[{"aliases": ["src/calendarkit/compat.py", "compat.py"], "kind": "file", "value": "src/calendarkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/quota_3.csv", "quota_3.csv"], "kind": "file", "value": "data/quota_3.csv"}, {"alias...
[{"answers": ["src/calendarkit/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-s1-00862
bucketkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] merge_probe_once lets defaults overwrite explicit overrides. It lives in src/bucketkit/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 bucketkit layout. [tool_use k...
98,185
3,072
115
140
false
synth:2386831286270149344@1.89833
synth
[{"aliases": ["src/bucketkit/compat.py", "compat.py"], "kind": "file", "value": "src/bucketkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/trace_5.csv", "trace_5.csv"], "kind": "file", "value": "data/trace_5.csv"}, {"aliases":...
[{"answers": ["src/bucketkit/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-s1-00862-twin
bucketkit
ctx_compress-s1-00862
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_probe_once lets defaults overwrite explicit overrides. It lives in src/bucketkit/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 bucketkit layout. [tool_use k...
62,931
3,072
81
140
false
synth:2386831286270149344@1.89833
synth
[{"aliases": ["src/bucketkit/compat.py", "compat.py"], "kind": "file", "value": "src/bucketkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/trace_5.csv", "trace_5.csv"], "kind": "file", "value": "data/trace_5.csv"}, {"aliases":...
[{"answers": ["src/bucketkit/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 mo...
ctx_compress-s1-00863
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] truncate_stem produces doubled dashes for consecutive separators. It lives in src/lexemeworks/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 lexemeworks layout. [too...
98,212
3,072
117
126
false
synth:1852131551985575497@1.71635
synth
[{"aliases": ["src/lexemeworks/rules.py", "rules.py"], "kind": "file", "value": "src/lexemeworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/lexemeworks/format.py", "format.py"], "kind": "file", "value": "src/lexemeworks/format.py"...
[{"answers": ["src/lexemeworks/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-s1-00863-twin
lexemeworks
ctx_compress-s1-00863
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] truncate_stem produces doubled dashes for consecutive separators. It lives in src/lexemeworks/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 lexemeworks layout. [too...
48,916
2,445
73
126
false
synth:1852131551985575497@1.71635
synth
[{"aliases": ["src/lexemeworks/rules.py", "rules.py"], "kind": "file", "value": "src/lexemeworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/lexemeworks/format.py", "format.py"], "kind": "file", "value": "src/lexemeworks/format.py"...
[{"answers": ["src/lexemeworks/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-00864
shiftworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_slot_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/shiftworks/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 shiftw...
86,136
3,072
112
131
false
synth:8754691432690474511@2.27338
synth
[{"aliases": ["src/shiftworks/report.py", "report.py"], "kind": "file", "value": "src/shiftworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/shiftworks/core.py", "core.py"], "kind": "file", "value": "src/shiftworks/core.py"}, {...
[{"answers": ["src/shiftworks/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-s1-00864-twin
shiftworks
ctx_compress-s1-00864
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_slot_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/shiftworks/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 shiftw...
45,034
2,251
62
131
false
synth:8754691432690474511@2.27338
synth
[{"aliases": ["src/shiftworks/report.py", "report.py"], "kind": "file", "value": "src/shiftworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/shiftworks/core.py", "core.py"], "kind": "file", "value": "src/shiftworks/core.py"}, {...
[{"answers": ["src/shiftworks/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-00865
shiftworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_shift produces doubled dashes for consecutive separators. It lives in src/shiftworks/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 shiftworks layout. [tool_use ki...
63,589
3,072
93
106
false
synth:5722739486569378901@1.46542
synth
[{"aliases": ["src/shiftworks/io.py", "io.py"], "kind": "file", "value": "src/shiftworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/shiftworks/audit.py", "audit.py"], "kind": "file", "value": "src/shiftworks/audit.py"}, {"aliases": ["src/shif...
[{"answers": ["src/shiftworks/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-00865-twin
shiftworks
ctx_compress-s1-00865
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_shift produces doubled dashes for consecutive separators. It lives in src/shiftworks/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 shiftworks layout. [tool_use ki...
39,449
1,972
61
106
false
synth:5722739486569378901@1.46542
synth
[{"aliases": ["src/shiftworks/io.py", "io.py"], "kind": "file", "value": "src/shiftworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/shiftworks/audit.py", "audit.py"], "kind": "file", "value": "src/shiftworks/audit.py"}, {"aliases": ["src/shif...
[{"answers": ["src/shiftworks/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-00866
glyphflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_lexeme_window ignores the cap, so late attempts wait for minutes. It lives in src/glyphflow/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 glyphflow layout. [...
39,493
1,974
64
73
false
synth:1475600420878984649@1
synth
[{"aliases": ["src/glyphflow/limits.py", "limits.py"], "kind": "file", "value": "src/glyphflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/glyphflow/rules.py", "rules.py"], "kind": "file", "value": "src/glyphflow/rules.py"}, {"...
[{"answers": ["src/glyphflow/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-s1-00866-twin
glyphflow
ctx_compress-s1-00866
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_lexeme_window ignores the cap, so late attempts wait for minutes. It lives in src/glyphflow/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 glyphflow layout. [...
18,711
1,000
42
73
false
synth:1475600420878984649@1
synth
[{"aliases": ["src/glyphflow/limits.py", "limits.py"], "kind": "file", "value": "src/glyphflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/glyphflow/rules.py", "rules.py"], "kind": "file", "value": "src/glyphflow/rules.py"}, {"...
[{"answers": ["src/glyphflow/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-s1-00867
slotflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_cadence raises QuotaDriftError for bytes input in strict mode instead of decoding it. It lives in src/slotflow/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 slo...
38,093
1,904
68
85
false
synth:4430686152884796468@1
synth
[{"aliases": ["src/slotflow/core.py", "core.py"], "kind": "file", "value": "src/slotflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/slotflow/compat.py", "compat.py"], "kind": "file", "value": "src/slotflow/compat.py"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["cli.py", "src/slotflow/cli.py"], "kin...
ctx_compress-s1-00867-twin
slotflow
ctx_compress-s1-00867
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_cadence raises QuotaDriftError for bytes input in strict mode instead of decoding it. It lives in src/slotflow/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 slo...
22,814
1,140
48
85
false
synth:4430686152884796468@1
synth
[{"aliases": ["src/slotflow/core.py", "core.py"], "kind": "file", "value": "src/slotflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/slotflow/compat.py", "compat.py"], "kind": "file", "value": "src/slotflow/compat.py"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["audit.py", "src/slotflow/audit.py"], ...
ctx_compress-s1-00868
seriesworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] merge_counter rejects a range where lo equals hi, which is a valid single point range. It lives in src/seriesworks/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 serie...
93,191
3,072
136
151
false
synth:5718316991341500273@2.14295
synth
[{"aliases": ["src/seriesworks/core.py", "core.py"], "kind": "file", "value": "src/seriesworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/span_6.csv", "span_6.csv"], "kind": "file", "value": "data/span_6.csv"}, {"aliases": ["src/seri...
[{"answers": ["src/seriesworks/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-s1-00868-twin
seriesworks
ctx_compress-s1-00868
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] merge_counter rejects a range where lo equals hi, which is a valid single point range. It lives in src/seriesworks/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 serie...
52,133
2,606
72
151
false
synth:5718316991341500273@2.14295
synth
[{"aliases": ["src/seriesworks/core.py", "core.py"], "kind": "file", "value": "src/seriesworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/span_6.csv", "span_6.csv"], "kind": "file", "value": "data/span_6.csv"}, {"aliases": ["src/seri...
[{"answers": ["src/seriesworks/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-s1-00869
binforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_reorder_once drops override values of 0, False and empty string. It lives in src/binforge/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 binforge layout. [tool...
67,273
3,072
89
114
false
synth:8326660083470506547@2.22578
synth
[{"aliases": ["src/binforge/report.py", "report.py"], "kind": "file", "value": "src/binforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases"...
[{"answers": ["src/binforge/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-00869-twin
binforge
ctx_compress-s1-00869
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_reorder_once drops override values of 0, False and empty string. It lives in src/binforge/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 binforge layout. [tool...
40,352
2,017
53
114
false
synth:8326660083470506547@2.22578
synth
[{"aliases": ["src/binforge/report.py", "report.py"], "kind": "file", "value": "src/binforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"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-s1-00870
principalworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_grant raises KeyError for a missing key instead of returning the default. It lives in src/principalworks/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 principa...
27,223
1,361
56
71
false
synth:2837342875515723330@1
synth
[{"aliases": ["src/principalworks/store.py", "store.py"], "kind": "file", "value": "src/principalworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/session_5.csv", "session_5.csv"], "kind": "file", "value": "data/session_5.csv"}, {...
[{"answers": ["src/principalworks/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-00870-twin
principalworks
ctx_compress-s1-00870
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_grant raises KeyError for a missing key instead of returning the default. It lives in src/principalworks/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 principa...
17,301
1,000
38
71
false
synth:2837342875515723330@1
synth
[{"aliases": ["src/principalworks/store.py", "store.py"], "kind": "file", "value": "src/principalworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/session_5.csv", "session_5.csv"], "kind": "file", "value": "data/session_5.csv"}, {...
[{"answers": ["src/principalworks/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-00871
zoneworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 returns the upper bound for every value instead of clamping into the range. It lives in src/zoneworks/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 zoneworks ...
39,834
1,991
71
84
false
synth:5992394880580521614@1
synth
[{"aliases": ["src/zoneworks/io.py", "io.py"], "kind": "file", "value": "src/zoneworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/zoneworks/format.py", "format.py"], "kind": "file", "value": "src/zoneworks/format.py"}, {"aliases": ["docs/chan...
[{"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": ["beacon_1.csv", "data/beacon_1.csv"], ...
ctx_compress-s1-00871-twin
zoneworks
ctx_compress-s1-00871
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 returns the upper bound for every value instead of clamping into the range. It lives in src/zoneworks/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 zoneworks ...
20,886
1,044
43
84
false
synth:5992394880580521614@1
synth
[{"aliases": ["src/zoneworks/io.py", "io.py"], "kind": "file", "value": "src/zoneworks/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/zoneworks/format.py", "format.py"], "kind": "file", "value": "src/zoneworks/format.py"}, {"aliases": ["docs/chan...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["compat.py", "src/zoneworks/compat.py"...
ctx_compress-s1-00872
skudesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_pallet keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/skudesk/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 skudesk layout. [...
49,833
2,491
100
107
false
synth:1964203397368112886@1.0653
synth
[{"aliases": ["src/skudesk/limits.py", "limits.py"], "kind": "file", "value": "src/skudesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/pallet_0.csv", "pallet_0.csv"], "kind": "file", "value": "data/pallet_0.csv"}, {"aliases": ...
[{"answers": ["src/skudesk/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-00872-twin
skudesk
ctx_compress-s1-00872
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] age_pallet keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/skudesk/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 skudesk layout. [...
25,003
1,250
49
107
false
synth:1964203397368112886@1.0653
synth
[{"aliases": ["src/skudesk/limits.py", "limits.py"], "kind": "file", "value": "src/skudesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/pallet_0.csv", "pallet_0.csv"], "kind": "file", "value": "data/pallet_0.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/skudesk/store.py", "store.py"], "...
ctx_compress-s1-00873
footnoteledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 averages one value too many once the window is full. It lives in src/footnoteledger/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 footnoteledger layo...
77,467
3,072
105
120
false
synth:6290450663707448144@1.7212
synth
[{"aliases": ["src/footnoteledger/limits.py", "limits.py"], "kind": "file", "value": "src/footnoteledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/footnoteledger/io.py", "io.py"], "kind": "file", "value": "src/footnoteledger/...
[{"answers": ["src/footnoteledger/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent...
ctx_compress-s1-00873-twin
footnoteledger
ctx_compress-s1-00873
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 averages one value too many once the window is full. It lives in src/footnoteledger/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 footnoteledger layo...
30,445
1,522
57
120
false
synth:6290450663707448144@1.7212
synth
[{"aliases": ["src/footnoteledger/limits.py", "limits.py"], "kind": "file", "value": "src/footnoteledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/footnoteledger/io.py", "io.py"], "kind": "file", "value": "src/footnoteledger/...
[{"answers": ["src/footnoteledger/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent...
ctx_compress-s1-00874
stockworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_warehouse keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/stockworks/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 stockw...
71,396
3,072
97
106
false
synth:8292729308332778468@1.26645
synth
[{"aliases": ["src/stockworks/compat.py", "compat.py"], "kind": "file", "value": "src/stockworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["src/stockworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-00874-twin
stockworks
ctx_compress-s1-00874
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_warehouse keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/stockworks/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 stockw...
32,420
1,621
51
106
false
synth:8292729308332778468@1.26645
synth
[{"aliases": ["src/stockworks/compat.py", "compat.py"], "kind": "file", "value": "src/stockworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"alia...
[{"answers": ["src/stockworks/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-s1-00875
tokenforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_claim raises PrincipalShapeError for bytes input in strict mode instead of decoding it. It lives in src/tokenforge/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the to...
18,251
1,000
54
59
false
synth:3228173190120213770@1
synth
[{"aliases": ["src/tokenforge/io.py", "io.py"], "kind": "file", "value": "src/tokenforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/tokenforge/report.py", "report.py"], "kind": "file", "value": "src/tokenforge/report.py"}, {"aliases": ["data/...
[{"answers": ["src/tokenforge/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-00875-twin
tokenforge
ctx_compress-s1-00875
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_claim raises PrincipalShapeError for bytes input in strict mode instead of decoding it. It lives in src/tokenforge/io.py; `python -m pytest tests/test_io.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the to...
9,683
1,000
32
59
false
synth:3228173190120213770@1
synth
[{"aliases": ["src/tokenforge/io.py", "io.py"], "kind": "file", "value": "src/tokenforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/tokenforge/report.py", "report.py"], "kind": "file", "value": "src/tokenforge/report.py"}, {"aliases": ["data/...
[{"answers": ["src/tokenforge/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rece...
ctx_compress-s1-00876
windowledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_slot_once raises ValueError when a value itself contains an equals sign. It lives in src/windowledger/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 windowl...
54,852
2,742
82
103
false
synth:4392370088734622540@1.2598
synth
[{"aliases": ["src/windowledger/compat.py", "compat.py"], "kind": "file", "value": "src/windowledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/windowledger/store.py", "store.py"], "kind": "file", "value": "src/windowledger/st...
[{"answers": ["src/windowledger/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-s1-00876-twin
windowledger
ctx_compress-s1-00876
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_slot_once raises ValueError when a value itself contains an equals sign. It lives in src/windowledger/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 windowl...
31,157
1,557
50
103
false
synth:4392370088734622540@1.2598
synth
[{"aliases": ["src/windowledger/compat.py", "compat.py"], "kind": "file", "value": "src/windowledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/windowledger/store.py", "store.py"], "kind": "file", "value": "src/windowledger/st...
[{"answers": ["src/windowledger/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-s1-00877
calendarflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_cadence_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/calendarflow/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 c...
48,603
2,430
69
82
false
synth:8805564126084112057@1
synth
[{"aliases": ["src/calendarflow/format.py", "format.py"], "kind": "file", "value": "src/calendarflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/calendar_0.csv", "calendar_0.csv"], "kind": "file", "value": "data/calendar_0.csv...
[{"answers": ["src/calendarflow/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-00877-twin
calendarflow
ctx_compress-s1-00877
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_cadence_strict keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/calendarflow/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 c...
23,900
1,195
41
82
false
synth:8805564126084112057@1
synth
[{"aliases": ["src/calendarflow/format.py", "format.py"], "kind": "file", "value": "src/calendarflow/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/calendar_0.csv", "calendar_0.csv"], "kind": "file", "value": "data/calendar_0.csv...
[{"answers": ["src/calendarflow/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-00878
principalledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] mint_claim drops override values of 0, False and empty string. It lives in src/principalledger/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 principalledger layou...
43,647
2,182
85
100
false
synth:7079390572663856963@1
synth
[{"aliases": ["src/principalledger/limits.py", "limits.py"], "kind": "file", "value": "src/principalledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/grant_0.csv", "grant_0.csv"], "kind": "file", "value": "data/grant_0.csv"},...
[{"answers": ["src/principalledger/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 agen...
ctx_compress-s1-00878-twin
principalledger
ctx_compress-s1-00878
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 drops override values of 0, False and empty string. It lives in src/principalledger/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 principalledger layou...
19,713
1,000
49
100
false
synth:7079390572663856963@1
synth
[{"aliases": ["src/principalledger/limits.py", "limits.py"], "kind": "file", "value": "src/principalledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/grant_0.csv", "grant_0.csv"], "kind": "file", "value": "data/grant_0.csv"},...
[{"answers": ["src/principalledger/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 agen...
ctx_compress-s1-00879
cadencekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_roster lets defaults overwrite explicit overrides. It lives in src/cadencekit/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 cadencekit layout. [tool_use kind=s...
42,394
2,119
79
96
false
synth:6517248183801959806@1
synth
[{"aliases": ["src/cadencekit/core.py", "core.py"], "kind": "file", "value": "src/cadencekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/calendar_0.csv", "calendar_0.csv"], "kind": "file", "value": "data/calendar_0.csv"}, {"aliases": ...
[{"answers": ["src/cadencekit/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-s1-00879-twin
cadencekit
ctx_compress-s1-00879
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_roster lets defaults overwrite explicit overrides. It lives in src/cadencekit/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 cadencekit layout. [tool_use kind=s...
23,533
1,176
53
96
false
synth:6517248183801959806@1
synth
[{"aliases": ["src/cadencekit/core.py", "core.py"], "kind": "file", "value": "src/cadencekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/calendar_0.csv", "calendar_0.csv"], "kind": "file", "value": "data/calendar_0.csv"}, {"aliases": ...
[{"answers": ["src/cadencekit/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-s1-00880
glyphworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_caption_window returns one item too few on every page. It lives in src/glyphworks/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 glyphworks layout. [tool...
51,378
2,568
91
114
false
synth:2931766648081081952@1.14878
synth
[{"aliases": ["src/glyphworks/report.py", "report.py"], "kind": "file", "value": "src/glyphworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/chunk_2.csv", "chunk_2.csv"], "kind": "file", "value": "data/chunk_2.csv"}, {"aliases...
[{"answers": ["src/glyphworks/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-s1-00880-twin
glyphworks
ctx_compress-s1-00880
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_caption_window returns one item too few on every page. It lives in src/glyphworks/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 glyphworks layout. [tool...
30,833
1,541
65
114
false
synth:2931766648081081952@1.14878
synth
[{"aliases": ["src/glyphworks/report.py", "report.py"], "kind": "file", "value": "src/glyphworks/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/chunk_2.csv", "chunk_2.csv"], "kind": "file", "value": "data/chunk_2.csv"}, {"aliases...
[{"answers": ["src/glyphworks/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-s1-00881
seriesdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/seriesdesk/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 seriesdesk la...
75,939
3,072
90
101
false
synth:3735136625985089739@1.94565
synth
[{"aliases": ["src/seriesdesk/format.py", "format.py"], "kind": "file", "value": "src/seriesdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/span_2.csv", "span_2.csv"], "kind": "file", "value": "data/span_2.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/seriesdesk/format.py...
ctx_compress-s1-00881-twin
seriesdesk
ctx_compress-s1-00881
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] clip_bucket keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/seriesdesk/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 seriesdesk la...
51,980
2,599
60
101
false
synth:3735136625985089739@1.94565
synth
[{"aliases": ["src/seriesdesk/format.py", "format.py"], "kind": "file", "value": "src/seriesdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/span_2.csv", "span_2.csv"], "kind": "file", "value": "data/span_2.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_cli.py", "tests/test_cli.py"], "...
ctx_compress-s1-00882
stemflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_caption raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/stemflow/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 stem...
97,616
3,072
101
108
false
synth:2566661501270306313@1.62423
synth
[{"aliases": ["src/stemflow/rules.py", "rules.py"], "kind": "file", "value": "src/stemflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemflow/schema.py", "schema.py"], "kind": "file", "value": "src/stemflow/schema.py"}, {"aliases...
[{"answers": ["src/stemflow/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-s1-00882-twin
stemflow
ctx_compress-s1-00882
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_caption raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/stemflow/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 stem...
50,097
2,504
61
108
false
synth:2566661501270306313@1.62423
synth
[{"aliases": ["src/stemflow/rules.py", "rules.py"], "kind": "file", "value": "src/stemflow/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/stemflow/schema.py", "schema.py"], "kind": "file", "value": "src/stemflow/schema.py"}, {"aliases...
[{"answers": ["src/stemflow/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-s1-00883
captionledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_cue leaves a trailing dash when the title ends in punctuation. It lives in src/captionledger/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 captionledger layout. ...
101,685
3,072
110
127
false
synth:1369214177429138010@1.89128
synth
[{"aliases": ["src/captionledger/core.py", "core.py"], "kind": "file", "value": "src/captionledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/lexeme_9.csv", "lexeme_9.csv"], "kind": "file", "value": "data/lexeme_9.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_rules.py", "tests/test_rules.py"...
ctx_compress-s1-00883-twin
captionledger
ctx_compress-s1-00883
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] wrap_cue leaves a trailing dash when the title ends in punctuation. It lives in src/captionledger/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 captionledger layout. ...
58,450
2,922
68
127
false
synth:1369214177429138010@1.89128
synth
[{"aliases": ["src/captionledger/core.py", "core.py"], "kind": "file", "value": "src/captionledger/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/lexeme_9.csv", "lexeme_9.csv"], "kind": "file", "value": "data/lexeme_9.csv"}, {"aliases": ...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/glyph_6.csv", "glyph_6.csv"], "k...
ctx_compress-s1-00884
ledgerkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_receipt_strict raises ValueError when a value itself contains an equals sign. It lives in src/ledgerkit/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 ledgerk...
50,327
2,516
76
81
false
synth:7784784685642295661@1
synth
[{"aliases": ["src/ledgerkit/compat.py", "compat.py"], "kind": "file", "value": "src/ledgerkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/receipt_2.csv", "receipt_2.csv"], "kind": "file", "value": "data/receipt_2.csv"}, {"ali...
[{"answers": ["src/ledgerkit/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 mo...
ctx_compress-s1-00884-twin
ledgerkit
ctx_compress-s1-00884
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] roll_receipt_strict raises ValueError when a value itself contains an equals sign. It lives in src/ledgerkit/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 ledgerk...
34,067
1,703
42
81
false
synth:7784784685642295661@1
synth
[{"aliases": ["src/ledgerkit/compat.py", "compat.py"], "kind": "file", "value": "src/ledgerkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/receipt_2.csv", "receipt_2.csv"], "kind": "file", "value": "data/receipt_2.csv"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_audit.py", "tests/test_audit.py"...
ctx_compress-s1-00885
slugdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_slug raises ChunkDriftError for bytes input in strict mode instead of decoding it. It lives in src/slugdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping th...
104,282
3,072
121
130
false
synth:1906304004814316182@1.91227
synth
[{"aliases": ["src/slugdesk/format.py", "format.py"], "kind": "file", "value": "src/slugdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slugdesk/io.py", "io.py"], "kind": "file", "value": "src/slugdesk/io.py"}, {"aliases": ["d...
[{"answers": ["src/slugdesk/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-00885-twin
slugdesk
ctx_compress-s1-00885
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] normalise_slug raises ChunkDriftError for bytes input in strict mode instead of decoding it. It lives in src/slugdesk/format.py; `python -m pytest tests/test_format.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping th...
46,402
2,320
69
130
false
synth:1906304004814316182@1.91227
synth
[{"aliases": ["src/slugdesk/format.py", "format.py"], "kind": "file", "value": "src/slugdesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/slugdesk/io.py", "io.py"], "kind": "file", "value": "src/slugdesk/io.py"}, {"aliases": ["d...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/stem_1.csv", "stem_1.csv"], "kin...
ctx_compress-s1-00886
receiptdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window skips the first page; page 1 returns what should be page 2. It lives in src/receiptdesk/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 receip...
101,439
3,072
111
140
false
synth:4063814031974798435@1.97365
synth
[{"aliases": ["src/receiptdesk/report.py", "report.py"], "kind": "file", "value": "src/receiptdesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/receiptdesk/schema.py", "schema.py"], "kind": "file", "value": "src/receiptdesk/sche...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/invoice_13.csv", "invoice_13.csv...
ctx_compress-s1-00886-twin
receiptdesk
ctx_compress-s1-00886
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window skips the first page; page 1 returns what should be page 2. It lives in src/receiptdesk/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 receip...
53,687
2,684
65
140
false
synth:4063814031974798435@1.97365
synth
[{"aliases": ["src/receiptdesk/report.py", "report.py"], "kind": "file", "value": "src/receiptdesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/receiptdesk/schema.py", "schema.py"], "kind": "file", "value": "src/receiptdesk/sche...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00887
captionforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_snippet_window keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/captionforge/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 caption...
65,341
3,072
99
118
false
synth:4588390057211686074@1.84522
synth
[{"aliases": ["src/captionforge/cli.py", "cli.py"], "kind": "file", "value": "src/captionforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/caption_1.csv", "caption_1.csv"], "kind": "file", "value": "data/caption_1.csv"}, {"aliases": ["src...
[{"answers": ["src/captionforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-00887-twin
captionforge
ctx_compress-s1-00887
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/captionforge/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 caption...
45,380
2,269
63
118
false
synth:4588390057211686074@1.84522
synth
[{"aliases": ["src/captionforge/cli.py", "cli.py"], "kind": "file", "value": "src/captionforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/caption_1.csv", "caption_1.csv"], "kind": "file", "value": "data/caption_1.csv"}, {"aliases": ["src...
[{"answers": ["src/captionforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-00888
chunkkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_cue_window drops override values of 0, False and empty string. It lives in src/chunkkit/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 chunkkit layout. [...
20,873
1,043
44
47
false
synth:8235196899728828504@1
synth
[{"aliases": ["src/chunkkit/compat.py", "compat.py"], "kind": "file", "value": "src/chunkkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/caption_4.csv", "caption_4.csv"], "kind": "file", "value": "data/caption_4.csv"}, {"alias...
[{"answers": ["tests/test_compat.py", "test_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-s1-00888-twin
chunkkit
ctx_compress-s1-00888
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_cue_window drops override values of 0, False and empty string. It lives in src/chunkkit/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 chunkkit layout. [...
11,493
1,000
28
47
false
synth:8235196899728828504@1
synth
[{"aliases": ["src/chunkkit/compat.py", "compat.py"], "kind": "file", "value": "src/chunkkit/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/caption_4.csv", "caption_4.csv"], "kind": "file", "value": "data/caption_4.csv"}, {"alias...
[{"answers": ["src/chunkkit/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 mos...
ctx_compress-s1-00889
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] tokenise_snippet drops override values of 0, False and empty string. It lives in src/captionworks/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 captionworks layou...
42,722
2,136
72
85
false
synth:5717991911866018707@1
synth
[{"aliases": ["src/captionworks/format.py", "format.py"], "kind": "file", "value": "src/captionworks/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/captionworks/audit.py", "audit.py"], "kind": "file", "value": "src/captionworks/au...
[{"answers": ["src/captionworks/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-00889-twin
captionworks
ctx_compress-s1-00889
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_snippet drops override values of 0, False and empty string. It lives in src/captionworks/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 captionworks layou...
20,694
1,034
48
85
false
synth:5717991911866018707@1
synth
[{"aliases": ["src/captionworks/format.py", "format.py"], "kind": "file", "value": "src/captionworks/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/captionworks/audit.py", "audit.py"], "kind": "file", "value": "src/captionworks/au...
[{"answers": ["src/captionworks/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-00890
zonedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_bearing_all keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/zonedesk/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 zonedes...
40,776
2,038
82
101
false
synth:2979620761060679657@1
synth
[{"aliases": ["src/zonedesk/core.py", "core.py"], "kind": "file", "value": "src/zonedesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/zonedesk/io.py", "io.py"], "kind": "file", "value": "src/zonedesk/io.py"}, {"aliases": ["src/zonedesk/...
[{"answers": ["src/zonedesk/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-00890-twin
zonedesk
ctx_compress-s1-00890
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_bearing_all keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/zonedesk/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 zonedes...
23,721
1,186
48
101
false
synth:2979620761060679657@1
synth
[{"aliases": ["src/zonedesk/core.py", "core.py"], "kind": "file", "value": "src/zonedesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/zonedesk/io.py", "io.py"], "kind": "file", "value": "src/zonedesk/io.py"}, {"aliases": ["src/zonedesk/...
[{"answers": ["src/zonedesk/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 re...
ctx_compress-s1-00891
windowkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_calendar double counts carried amounts when the period closes on a weekend. It lives in src/windowkit/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 windowkit ...
45,812
2,290
69
74
false
synth:8847621939507590451@1
synth
[{"aliases": ["src/windowkit/audit.py", "audit.py"], "kind": "file", "value": "src/windowkit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/shift_2.csv", "shift_2.csv"], "kind": "file", "value": "data/shift_2.csv"}, {"aliases": ["src...
[{"answers": ["src/windowkit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-00891-twin
windowkit
ctx_compress-s1-00891
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] split_calendar double counts carried amounts when the period closes on a weekend. It lives in src/windowkit/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 windowkit ...
20,470
1,023
37
74
false
synth:8847621939507590451@1
synth
[{"aliases": ["src/windowkit/audit.py", "audit.py"], "kind": "file", "value": "src/windowkit/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/shift_2.csv", "shift_2.csv"], "kind": "file", "value": "data/shift_2.csv"}, {"aliases": ["src...
[{"answers": ["src/windowkit/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-00892
counterforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] downsample_series_batch produces doubled dashes for consecutive separators. It lives in src/counterforge/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 counterforge ...
87,744
3,072
97
112
false
synth:3428504483024867606@1.859
synth
[{"aliases": ["src/counterforge/store.py", "store.py"], "kind": "file", "value": "src/counterforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/counterforge/tables.py", "tables.py"], "kind": "file", "value": "src/counterforge/tables...
[{"answers": ["src/counterforge/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-00892-twin
counterforge
ctx_compress-s1-00892
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] downsample_series_batch produces doubled dashes for consecutive separators. It lives in src/counterforge/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 counterforge ...
49,413
2,470
63
112
false
synth:3428504483024867606@1.859
synth
[{"aliases": ["src/counterforge/store.py", "store.py"], "kind": "file", "value": "src/counterforge/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/counterforge/tables.py", "tables.py"], "kind": "file", "value": "src/counterforge/tables...
[{"answers": ["src/counterforge/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-00893
bookingflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_booking raises ValueError when a value itself contains an equals sign. It lives in src/bookingflow/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 bookingflow layo...
40,851
2,042
76
95
false
synth:8832059982224466978@1.34197
synth
[{"aliases": ["src/bookingflow/cli.py", "cli.py"], "kind": "file", "value": "src/bookingflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/shi...
[{"answers": ["src/bookingflow/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-s1-00893-twin
bookingflow
ctx_compress-s1-00893
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_booking raises ValueError when a value itself contains an equals sign. It lives in src/bookingflow/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 bookingflow layo...
25,152
1,257
47
95
false
synth:8832059982224466978@1.34197
synth
[{"aliases": ["src/bookingflow/cli.py", "cli.py"], "kind": "file", "value": "src/bookingflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ["data/shi...
[{"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-s1-00894
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 averages one value too many once the window is full. It lives in src/glyphkit/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 glyphkit layout. [tool_u...
52,411
2,620
91
104
false
synth:3310950064503589528@1.09597
synth
[{"aliases": ["src/glyphkit/core.py", "core.py"], "kind": "file", "value": "src/glyphkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/glyphkit/limits.py", "limits.py"], "kind": "file", "value": "src/glyphkit/limits.py"}, {"aliases": ["d...
[{"answers": ["src/glyphkit/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most re...
ctx_compress-s1-00894-twin
glyphkit
ctx_compress-s1-00894
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 averages one value too many once the window is full. It lives in src/glyphkit/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 glyphkit layout. [tool_u...
28,506
1,425
55
104
false
synth:3310950064503589528@1.09597
synth
[{"aliases": ["src/glyphkit/core.py", "core.py"], "kind": "file", "value": "src/glyphkit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/glyphkit/limits.py", "limits.py"], "kind": "file", "value": "src/glyphkit/limits.py"}, {"aliases": ["d...
[{"answers": ["src/glyphkit/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 re...
ctx_compress-s1-00895
clauseworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index misses keys that differ only in case. It lives in src/clauseworks/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 clauseworks layout. [tool_use kind=sear...
72,844
3,072
89
104
false
synth:5072994741651161864@1.84028
synth
[{"aliases": ["src/clauseworks/core.py", "core.py"], "kind": "file", "value": "src/clauseworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/clauseworks/format.py", "format.py"], "kind": "file", "value": "src/clauseworks/format.py"}, {"a...
[{"answers": ["src/clauseworks/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-s1-00895-twin
clauseworks
ctx_compress-s1-00895
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] paginate_index misses keys that differ only in case. It lives in src/clauseworks/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 clauseworks layout. [tool_use kind=sear...
37,969
1,898
53
104
false
synth:5072994741651161864@1.84028
synth
[{"aliases": ["src/clauseworks/core.py", "core.py"], "kind": "file", "value": "src/clauseworks/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/clauseworks/format.py", "format.py"], "kind": "file", "value": "src/clauseworks/format.py"}, {"a...
[{"answers": ["src/clauseworks/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-s1-00896
tokendesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_claim_once grows linearly instead of doubling. It lives in src/tokendesk/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 tokendesk layout. [tool_use kind=sea...
23,629
1,181
40
49
false
synth:3353757090040652931@1
synth
[{"aliases": ["src/tokendesk/format.py", "format.py"], "kind": "file", "value": "src/tokendesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tokendesk/report.py", "report.py"], "kind": "file", "value": "src/tokendesk/report.py"},...
[{"answers": ["src/tokendesk/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-00896-twin
tokendesk
ctx_compress-s1-00896
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] verify_claim_once grows linearly instead of doubling. It lives in src/tokendesk/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 tokendesk layout. [tool_use kind=sea...
8,931
1,000
25
49
false
synth:3353757090040652931@1
synth
[{"aliases": ["src/tokendesk/format.py", "format.py"], "kind": "file", "value": "src/tokendesk/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["src/tokendesk/report.py", "report.py"], "kind": "file", "value": "src/tokendesk/report.py"},...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["claim_4.csv", "data/claim_4.csv"], "k...
ctx_compress-s1-00897
footnoteworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_index_page raises ValueError when a value itself contains an equals sign. It lives in src/footnoteworks/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 footnot...
52,250
2,612
99
108
false
synth:8576113188220515162@1
synth
[{"aliases": ["src/footnoteworks/rules.py", "rules.py"], "kind": "file", "value": "src/footnoteworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/footnoteworks/tables.py", "tables.py"], "kind": "file", "value": "src/footnoteworks/ta...
[{"answers": ["src/footnoteworks/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 ...
ctx_compress-s1-00897-twin
footnoteworks
ctx_compress-s1-00897
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_index_page raises ValueError when a value itself contains an equals sign. It lives in src/footnoteworks/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 footnot...
22,025
1,101
49
108
false
synth:8576113188220515162@1
synth
[{"aliases": ["src/footnoteworks/rules.py", "rules.py"], "kind": "file", "value": "src/footnoteworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/footnoteworks/tables.py", "tables.py"], "kind": "file", "value": "src/footnoteworks/ta...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00898
corridorkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_beacon_page drops override values of 0, False and empty string. It lives in src/corridorkit/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 corridorkit ...
108,114
3,072
143
154
false
synth:5796748916583894095@2.14478
synth
[{"aliases": ["src/corridorkit/limits.py", "limits.py"], "kind": "file", "value": "src/corridorkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/beacon_7.csv", "beacon_7.csv"], "kind": "file", "value": "data/beacon_7.csv"}, {"al...
[{"answers": ["src/corridorkit/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-s1-00898-twin
corridorkit
ctx_compress-s1-00898
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_beacon_page drops override values of 0, False and empty string. It lives in src/corridorkit/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 corridorkit ...
60,438
3,021
87
154
false
synth:5796748916583894095@2.14478
synth
[{"aliases": ["src/corridorkit/limits.py", "limits.py"], "kind": "file", "value": "src/corridorkit/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/beacon_7.csv", "beacon_7.csv"], "kind": "file", "value": "data/beacon_7.csv"}, {"al...
[{"answers": ["src/corridorkit/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-s1-00899
bearingforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 leaves a trailing dash when the title ends in punctuation. It lives in src/bearingforge/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 bearingforge layout. [...
62,655
3,072
110
129
false
synth:8371283879323469924@1.7177
synth
[{"aliases": ["src/bearingforge/core.py", "core.py"], "kind": "file", "value": "src/bearingforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/bearingforge/store.py", "store.py"], "kind": "file", "value": "src/bearingforge/store.py"}, {"...
[{"answers": ["src/bearingforge/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-00899-twin
bearingforge
ctx_compress-s1-00899
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 leaves a trailing dash when the title ends in punctuation. It lives in src/bearingforge/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 bearingforge layout. [...
35,008
1,750
66
129
false
synth:8371283879323469924@1.7177
synth
[{"aliases": ["src/bearingforge/core.py", "core.py"], "kind": "file", "value": "src/bearingforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/bearingforge/store.py", "store.py"], "kind": "file", "value": "src/bearingforge/store.py"}, {"...
[{"answers": ["src/bearingforge/core.py", "core.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-00900
zoneforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_zone raises KeyError for a missing key instead of returning the default. It lives in src/zoneforge/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 zoneforge layo...
82,945
3,072
112
121
false
synth:7717814493386327568@1.55767
synth
[{"aliases": ["src/zoneforge/cli.py", "cli.py"], "kind": "file", "value": "src/zoneforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/zoneforge/core.py", "core.py"], "kind": "file", "value": "src/zoneforge/core.py"}, {"aliases": ["data/zone...
[{"answers": ["src/zoneforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rec...
ctx_compress-s1-00900-twin
zoneforge
ctx_compress-s1-00900
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_zone raises KeyError for a missing key instead of returning the default. It lives in src/zoneforge/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 zoneforge layo...
29,354
1,467
58
121
false
synth:7717814493386327568@1.55767
synth
[{"aliases": ["src/zoneforge/cli.py", "cli.py"], "kind": "file", "value": "src/zoneforge/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/zoneforge/core.py", "core.py"], "kind": "file", "value": "src/zoneforge/core.py"}, {"aliases": ["data/zone...
[{"answers": ["src/zoneforge/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-00901
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_tariff raises CreditDriftError for bytes input in strict mode instead of decoding it. It lives in src/receiptledger/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...
108,193
3,072
116
127
false
synth:4396458917391467534@2.261
synth
[{"aliases": ["src/receiptledger/store.py", "store.py"], "kind": "file", "value": "src/receiptledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/balance_1.csv", "balance_1.csv"], "kind": "file", "value": "data/balance_1.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["src/receiptledger/store.py", "store.p...
ctx_compress-s1-00901-twin
receiptledger
ctx_compress-s1-00901
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_tariff raises CreditDriftError for bytes input in strict mode instead of decoding it. It lives in src/receiptledger/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...
59,993
2,999
70
127
false
synth:4396458917391467534@2.261
synth
[{"aliases": ["src/receiptledger/store.py", "store.py"], "kind": "file", "value": "src/receiptledger/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/balance_1.csv", "balance_1.csv"], "kind": "file", "value": "data/balance_1.csv"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["changelog.md", "docs/changelog.md"], ...
ctx_compress-s1-00902
shiftflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_slot rejects a range where lo equals hi, which is a valid single point range. It lives in src/shiftflow/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 shiftfl...
67,536
3,072
96
109
false
synth:9120426930871001706@1.21453
synth
[{"aliases": ["src/shiftflow/store.py", "store.py"], "kind": "file", "value": "src/shiftflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/shiftflow/rules.py", "rules.py"], "kind": "file", "value": "src/shiftflow/rules.py"}, {"aliase...
[{"answers": ["src/shiftflow/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-00902-twin
shiftflow
ctx_compress-s1-00902
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_slot rejects a range where lo equals hi, which is a valid single point range. It lives in src/shiftflow/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 shiftfl...
36,194
1,809
58
109
false
synth:9120426930871001706@1.21453
synth
[{"aliases": ["src/shiftflow/store.py", "store.py"], "kind": "file", "value": "src/shiftflow/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/shiftflow/rules.py", "rules.py"], "kind": "file", "value": "src/shiftflow/rules.py"}, {"aliase...
[{"answers": ["src/shiftflow/store.py", "store.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most...
ctx_compress-s1-00903
sessiondesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope_batch averages one value too many once the window is full. It lives in src/sessiondesk/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 sessiondesk layout...
52,324
2,616
100
107
false
synth:1773588726475162801@1.10532
synth
[{"aliases": ["src/sessiondesk/rules.py", "rules.py"], "kind": "file", "value": "src/sessiondesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/policy_0.csv", "policy_0.csv"], "kind": "file", "value": "data/policy_0.csv"}, {"aliases"...
[{"answers": ["src/sessiondesk/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-s1-00903-twin
sessiondesk
ctx_compress-s1-00903
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_scope_batch averages one value too many once the window is full. It lives in src/sessiondesk/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 sessiondesk layout...
27,983
1,399
54
107
false
synth:1773588726475162801@1.10532
synth
[{"aliases": ["src/sessiondesk/rules.py", "rules.py"], "kind": "file", "value": "src/sessiondesk/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/policy_0.csv", "policy_0.csv"], "kind": "file", "value": "data/policy_0.csv"}, {"aliases"...
[{"answers": ["src/sessiondesk/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-00904
cadencekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_shift_page returns one item too few on every page. It lives in src/cadencekit/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 cadencekit layout. [tool_use kind=...
64,115
3,072
89
112
false
synth:1621959479411965952@1.29
synth
[{"aliases": ["src/cadencekit/core.py", "core.py"], "kind": "file", "value": "src/cadencekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/booking_1.csv", "booking_1.csv"], "kind": "file", "value": "data/booking_1.csv"}, {"aliases": ["d...
[{"answers": ["src/cadencekit/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-s1-00904-twin
cadencekit
ctx_compress-s1-00904
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_shift_page returns one item too few on every page. It lives in src/cadencekit/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 cadencekit layout. [tool_use kind=...
40,300
2,015
65
112
false
synth:1621959479411965952@1.29
synth
[{"aliases": ["src/cadencekit/core.py", "core.py"], "kind": "file", "value": "src/cadencekit/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/booking_1.csv", "booking_1.csv"], "kind": "file", "value": "data/booking_1.csv"}, {"aliases": ["d...
[{"answers": ["src/cadencekit/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 ...