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-s1-01056
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] balance_roster raises ValueError when a value itself contains an equals sign. 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 calendarf...
101,349
3,072
119
144
false
synth:7365096531595438456@2.14728
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": ["src/calendarflow/compat.py", "compat.py"], "kind": "file", "value": "src/calendarflow/...
[{"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-s1-01056-twin
calendarflow
ctx_compress-s1-01056
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] balance_roster raises ValueError when a value itself contains an equals sign. 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 calendarf...
60,766
3,038
79
144
false
synth:7365096531595438456@2.14728
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": ["src/calendarflow/compat.py", "compat.py"], "kind": "file", "value": "src/calendarflow/...
[{"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-s1-01057
shipmentforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_lot leaves a trailing dash when the title ends in punctuation. It lives in src/shipmentforge/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 shipmentforge layout. [to...
63,417
3,072
93
106
false
synth:8353630071860236555@2.34125
synth
[{"aliases": ["src/shipmentforge/io.py", "io.py"], "kind": "file", "value": "src/shipmentforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/shipmentforge/store.py", "store.py"], "kind": "file", "value": "src/shipmentforge/store.py"}, {"aliases"...
[{"answers": ["src/shipmentforge/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most r...
ctx_compress-s1-01057-twin
shipmentforge
ctx_compress-s1-01057
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] count_lot leaves a trailing dash when the title ends in punctuation. It lives in src/shipmentforge/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 shipmentforge layout. [to...
39,383
1,969
53
106
false
synth:8353630071860236555@2.34125
synth
[{"aliases": ["src/shipmentforge/io.py", "io.py"], "kind": "file", "value": "src/shipmentforge/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/shipmentforge/store.py", "store.py"], "kind": "file", "value": "src/shipmentforge/store.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": ["compat.py", "src/shipmentforge/compat...
ctx_compress-s1-01058
lexemeforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_stem_batch misses keys that differ only in case. It lives in src/lexemeforge/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 lexemeforge layout. [tool_use kind...
48,592
2,429
88
109
false
synth:2846714104632788486@1.0808
synth
[{"aliases": ["src/lexemeforge/audit.py", "audit.py"], "kind": "file", "value": "src/lexemeforge/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/lexemeforge/core.py", "core.py"], "kind": "file", "value": "src/lexemeforge/core.py"}, {"a...
[{"answers": ["src/lexemeforge/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-01058-twin
lexemeforge
ctx_compress-s1-01058
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_stem_batch misses keys that differ only in case. It lives in src/lexemeforge/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 lexemeforge layout. [tool_use kind...
32,245
1,612
54
109
false
synth:2846714104632788486@1.0808
synth
[{"aliases": ["src/lexemeforge/audit.py", "audit.py"], "kind": "file", "value": "src/lexemeforge/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/lexemeforge/core.py", "core.py"], "kind": "file", "value": "src/lexemeforge/core.py"}, {"a...
[{"answers": ["src/lexemeforge/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-01059
tokenworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_token lets defaults overwrite explicit overrides. It lives in src/tokenworks/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 tokenworks layout. [tool_use kin...
41,043
2,052
61
76
false
synth:5032412106112510289@1
synth
[{"aliases": ["src/tokenworks/compat.py", "compat.py"], "kind": "file", "value": "src/tokenworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tokenworks/report.py", "report.py"], "kind": "file", "value": "src/tokenworks/report.p...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/token_0.csv", "token_0.csv"], "k...
ctx_compress-s1-01059-twin
tokenworks
ctx_compress-s1-01059
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_token lets defaults overwrite explicit overrides. It lives in src/tokenworks/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 tokenworks layout. [tool_use kin...
33,209
1,660
41
76
false
synth:5032412106112510289@1
synth
[{"aliases": ["src/tokenworks/compat.py", "compat.py"], "kind": "file", "value": "src/tokenworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/tokenworks/report.py", "report.py"], "kind": "file", "value": "src/tokenworks/report.p...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/tokenworks/core.py"], ...
ctx_compress-s1-01060
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] roll_invoice_window returns one item too few on every page. It lives in src/receiptdesk/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 receiptdesk layout. [tool_us...
87,727
3,072
118
139
false
synth:7582131975720532789@2.24193
synth
[{"aliases": ["src/receiptdesk/limits.py", "limits.py"], "kind": "file", "value": "src/receiptdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/receiptdesk/cli.py", "cli.py"], "kind": "file", "value": "src/receiptdesk/cli.py"}, ...
[{"answers": ["src/receiptdesk/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-s1-01060-twin
receiptdesk
ctx_compress-s1-01060
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_invoice_window returns one item too few on every page. It lives in src/receiptdesk/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 receiptdesk layout. [tool_us...
54,039
2,701
80
139
false
synth:7582131975720532789@2.24193
synth
[{"aliases": ["src/receiptdesk/limits.py", "limits.py"], "kind": "file", "value": "src/receiptdesk/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/receiptdesk/cli.py", "cli.py"], "kind": "file", "value": "src/receiptdesk/cli.py"}, ...
[{"answers": ["src/receiptdesk/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-s1-01061
balanceforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_invoice_all skips the first page; page 1 returns what should be page 2. It lives in src/balanceforge/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 balancef...
93,719
3,072
105
122
false
synth:3712507883177176262@1.702
synth
[{"aliases": ["src/balanceforge/compat.py", "compat.py"], "kind": "file", "value": "src/balanceforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/balanceforge/limits.py", "limits.py"], "kind": "file", "value": "src/balanceforge/...
[{"answers": ["src/balanceforge/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-s1-01061-twin
balanceforge
ctx_compress-s1-01061
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_invoice_all skips the first page; page 1 returns what should be page 2. It lives in src/balanceforge/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 balancef...
60,652
3,032
65
122
false
synth:3712507883177176262@1.702
synth
[{"aliases": ["src/balanceforge/compat.py", "compat.py"], "kind": "file", "value": "src/balanceforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/balanceforge/limits.py", "limits.py"], "kind": "file", "value": "src/balanceforge/...
[{"answers": ["src/balanceforge/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-s1-01062
policyledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 produces doubled dashes for consecutive separators. It lives in src/policyledger/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 policyledger layout. [to...
86,291
3,072
106
121
false
synth:1346152967279394810@1.737
synth
[{"aliases": ["src/policyledger/audit.py", "audit.py"], "kind": "file", "value": "src/policyledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policyledger/tables.py", "tables.py"], "kind": "file", "value": "src/policyledger/tables...
[{"answers": ["src/policyledger/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01062-twin
policyledger
ctx_compress-s1-01062
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 produces doubled dashes for consecutive separators. It lives in src/policyledger/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 policyledger layout. [to...
50,441
2,522
68
121
false
synth:1346152967279394810@1.737
synth
[{"aliases": ["src/policyledger/audit.py", "audit.py"], "kind": "file", "value": "src/policyledger/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policyledger/tables.py", "tables.py"], "kind": "file", "value": "src/policyledger/tables...
[{"answers": ["src/policyledger/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's m...
ctx_compress-s1-01063
lexemekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/lexemekit/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 lexemekit layout. [tool_use...
48,474
2,423
82
95
false
synth:8126054201194135265@1.47745
synth
[{"aliases": ["src/lexemekit/cli.py", "cli.py"], "kind": "file", "value": "src/lexemekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/glyph_2.csv", "glyph_2.csv"], "kind": "file", "value": "data/glyph_2.csv"}, {"aliases": ["data/slug_1.csv...
[{"answers": ["src/lexemekit/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-s1-01063-twin
lexemekit
ctx_compress-s1-01063
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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/lexemekit/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 lexemekit layout. [tool_use...
32,581
1,629
53
95
false
synth:8126054201194135265@1.47745
synth
[{"aliases": ["src/lexemekit/cli.py", "cli.py"], "kind": "file", "value": "src/lexemekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/glyph_2.csv", "glyph_2.csv"], "kind": "file", "value": "data/glyph_2.csv"}, {"aliases": ["data/slug_1.csv...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/lexemekit/format.py"...
ctx_compress-s1-01064
reorderforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_warehouse_safe keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/reorderforge/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 ...
35,788
1,789
68
81
false
synth:5755051959629206009@1
synth
[{"aliases": ["src/reorderforge/report.py", "report.py"], "kind": "file", "value": "src/reorderforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/stock_3.csv", "stock_3.csv"], "kind": "file", "value": "data/stock_3.csv"}, {"ali...
[{"answers": ["src/reorderforge/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01064-twin
reorderforge
ctx_compress-s1-01064
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_warehouse_safe keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/reorderforge/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 ...
21,898
1,094
42
81
false
synth:5755051959629206009@1
synth
[{"aliases": ["src/reorderforge/report.py", "report.py"], "kind": "file", "value": "src/reorderforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/stock_3.csv", "stock_3.csv"], "kind": "file", "value": "data/stock_3.csv"}, {"ali...
[{"answers": ["src/reorderforge/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01065
revisionflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_section_safe averages one value too many once the window is full. It lives in src/revisionflow/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 revisionflow layout....
61,790
3,072
114
129
false
synth:6071521250638418782@2.34698
synth
[{"aliases": ["src/revisionflow/cli.py", "cli.py"], "kind": "file", "value": "src/revisionflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/revisionflow/tables.py", "tables.py"], "kind": "file", "value": "src/revisionflow/tables.py"}, {"ali...
[{"answers": ["src/revisionflow/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-s1-01065-twin
revisionflow
ctx_compress-s1-01065
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_section_safe averages one value too many once the window is full. It lives in src/revisionflow/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 revisionflow layout....
43,076
2,153
73
129
false
synth:6071521250638418782@2.34698
synth
[{"aliases": ["src/revisionflow/cli.py", "cli.py"], "kind": "file", "value": "src/revisionflow/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/revisionflow/tables.py", "tables.py"], "kind": "file", "value": "src/revisionflow/tables.py"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_store.py", "tests/test_store.py"...
ctx_compress-s1-01066
revisionledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_section averages one value too many once the window is full. It lives in src/revisionledger/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 revisionledger l...
74,645
3,072
112
121
false
synth:5207776155258181965@1.38645
synth
[{"aliases": ["src/revisionledger/compat.py", "compat.py"], "kind": "file", "value": "src/revisionledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/revisionledger/audit.py", "audit.py"], "kind": "file", "value": "src/revisionl...
[{"answers": ["src/revisionledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent...
ctx_compress-s1-01066-twin
revisionledger
ctx_compress-s1-01066
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] collate_section averages one value too many once the window is full. It lives in src/revisionledger/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 revisionledger l...
34,320
1,716
56
121
false
synth:5207776155258181965@1.38645
synth
[{"aliases": ["src/revisionledger/compat.py", "compat.py"], "kind": "file", "value": "src/revisionledger/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["src/revisionledger/audit.py", "audit.py"], "kind": "file", "value": "src/revisionl...
[{"answers": ["src/revisionledger/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent...
ctx_compress-s1-01067
refunddesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_ledger raises PayoutStateError for bytes input in strict mode instead of decoding it. It lives in src/refunddesk/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 th...
47,001
2,350
80
95
false
synth:3747310033006950184@1
synth
[{"aliases": ["src/refunddesk/core.py", "core.py"], "kind": "file", "value": "src/refunddesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/refunddesk/report.py", "report.py"], "kind": "file", "value": "src/refunddesk/report.py"}, {"alias...
[{"answers": ["src/refunddesk/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-s1-01067-twin
refunddesk
ctx_compress-s1-01067
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_ledger raises PayoutStateError for bytes input in strict mode instead of decoding it. It lives in src/refunddesk/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 th...
22,977
1,148
45
95
false
synth:3747310033006950184@1
synth
[{"aliases": ["src/refunddesk/core.py", "core.py"], "kind": "file", "value": "src/refunddesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/refunddesk/report.py", "report.py"], "kind": "file", "value": "src/refunddesk/report.py"}, {"alias...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/refunddesk/core.py"], ...
ctx_compress-s1-01068
draftforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_batch drops override values of 0, False and empty string. It lives in src/draftforge/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 draftforge layout. [...
35,793
1,789
68
81
false
synth:6716193405787748426@1
synth
[{"aliases": ["src/draftforge/core.py", "core.py"], "kind": "file", "value": "src/draftforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/draftforge/store.py", "store.py"], "kind": "file", "value": "src/draftforge/store.py"}, {"aliases"...
[{"answers": ["src/draftforge/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-s1-01068-twin
draftforge
ctx_compress-s1-01068
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_heading_batch drops override values of 0, False and empty string. It lives in src/draftforge/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 draftforge layout. [...
20,750
1,037
40
81
false
synth:6716193405787748426@1
synth
[{"aliases": ["src/draftforge/core.py", "core.py"], "kind": "file", "value": "src/draftforge/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/draftforge/store.py", "store.py"], "kind": "file", "value": "src/draftforge/store.py"}, {"aliases"...
[{"answers": ["src/draftforge/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-s1-01069
sessionworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_principal grows linearly instead of doubling. It lives in src/sessionworks/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 sessionworks layout. [tool_use kind=...
20,209
1,010
50
57
false
synth:4492505015444161297@1
synth
[{"aliases": ["src/sessionworks/store.py", "store.py"], "kind": "file", "value": "src/sessionworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/sessionworks/io.py", "io.py"], "kind": "file", "value": "src/sessionworks/io.py"}, {"ali...
[{"answers": ["src/sessionworks/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 m...
ctx_compress-s1-01069-twin
sessionworks
ctx_compress-s1-01069
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_principal grows linearly instead of doubling. It lives in src/sessionworks/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 sessionworks layout. [tool_use kind=...
14,357
1,000
30
57
false
synth:4492505015444161297@1
synth
[{"aliases": ["src/sessionworks/store.py", "store.py"], "kind": "file", "value": "src/sessionworks/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/sessionworks/io.py", "io.py"], "kind": "file", "value": "src/sessionworks/io.py"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["format.py", "src/sessionworks/format....
ctx_compress-s1-01070
headingdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] fold_margin double counts carried amounts when the period closes on a weekend. It lives in src/headingdesk/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 headingdesk layou...
70,827
3,072
86
103
false
synth:4623589081713566220@1.62587
synth
[{"aliases": ["src/headingdesk/io.py", "io.py"], "kind": "file", "value": "src/headingdesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/headingdesk/rules.py", "rules.py"], "kind": "file", "value": "src/headingdesk/rules.py"}, {"aliases": ["src/...
[{"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/draft_10.csv", "draft_10.csv"], ...
ctx_compress-s1-01070-twin
headingdesk
ctx_compress-s1-01070
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] fold_margin double counts carried amounts when the period closes on a weekend. It lives in src/headingdesk/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 headingdesk layou...
42,407
2,120
50
103
false
synth:4623589081713566220@1.62587
synth
[{"aliases": ["src/headingdesk/io.py", "io.py"], "kind": "file", "value": "src/headingdesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/headingdesk/rules.py", "rules.py"], "kind": "file", "value": "src/headingdesk/rules.py"}, {"aliases": ["src/...
[{"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/margin_8.csv", "margin_8.csv"], ...
ctx_compress-s1-01071
margindesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_clause lets defaults overwrite explicit overrides. It lives in src/margindesk/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 margindesk layout. [tool_use kind=s...
28,154
1,407
52
57
false
synth:6755701260735067829@1
synth
[{"aliases": ["src/margindesk/cli.py", "cli.py"], "kind": "file", "value": "src/margindesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/margindesk/report.py", "report.py"], "kind": "file", "value": "src/margindesk/report.py"}, {"aliases": [...
[{"answers": ["src/margindesk/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most re...
ctx_compress-s1-01071-twin
margindesk
ctx_compress-s1-01071
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] renumber_clause lets defaults overwrite explicit overrides. It lives in src/margindesk/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 margindesk layout. [tool_use kind=s...
11,791
1,000
29
57
false
synth:6755701260735067829@1
synth
[{"aliases": ["src/margindesk/cli.py", "cli.py"], "kind": "file", "value": "src/margindesk/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/margindesk/report.py", "report.py"], "kind": "file", "value": "src/margindesk/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": ["core.py", "src/margindesk/core.py"], ...
ctx_compress-s1-01072
cuekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_safe produces doubled dashes for consecutive separators. It lives in src/cuekit/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 cuekit layout. [tool_u...
82,872
3,072
96
117
false
synth:5434468920858578760@1.52032
synth
[{"aliases": ["src/cuekit/report.py", "report.py"], "kind": "file", "value": "src/cuekit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/cuekit/limits.py", "limits.py"], "kind": "file", "value": "src/cuekit/limits.py"}, {"aliases":...
[{"answers": ["src/cuekit/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most ...
ctx_compress-s1-01072-twin
cuekit
ctx_compress-s1-01072
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_safe produces doubled dashes for consecutive separators. It lives in src/cuekit/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 cuekit layout. [tool_u...
38,222
1,911
58
117
false
synth:5434468920858578760@1.52032
synth
[{"aliases": ["src/cuekit/report.py", "report.py"], "kind": "file", "value": "src/cuekit/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/cuekit/limits.py", "limits.py"], "kind": "file", "value": "src/cuekit/limits.py"}, {"aliases":...
[{"answers": ["src/cuekit/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most ...
ctx_compress-s1-01073
stemkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] truncate_chunk_once returns one item too few on every page. It lives in src/stemkit/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 stemkit layout. [tool_use kind=search ...
109,958
3,072
105
132
false
synth:1719299017853989327@2.08955
synth
[{"aliases": ["src/stemkit/cli.py", "cli.py"], "kind": "file", "value": "src/stemkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/chunk_2.csv", "chunk_2.csv"], "kind": "file", "value": "data/chunk_2.csv"}, {"aliases": ["data/cue_4.csv", "c...
[{"answers": ["src/stemkit/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 recen...
ctx_compress-s1-01073-twin
stemkit
ctx_compress-s1-01073
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] truncate_chunk_once returns one item too few on every page. It lives in src/stemkit/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 stemkit layout. [tool_use kind=search ...
67,295
3,072
79
132
false
synth:1719299017853989327@2.08955
synth
[{"aliases": ["src/stemkit/cli.py", "cli.py"], "kind": "file", "value": "src/stemkit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/chunk_2.csv", "chunk_2.csv"], "kind": "file", "value": "data/chunk_2.csv"}, {"aliases": ["data/cue_4.csv", "c...
[{"answers": ["src/stemkit/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 recen...
ctx_compress-s1-01074
cadenceworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_calendar_page returns the upper bound for every value instead of clamping into the range. It lives in src/cadenceworks/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 map...
79,144
3,072
105
124
false
synth:3101026652240151760@1.5709
synth
[{"aliases": ["src/cadenceworks/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/shift_7.csv", "shift_7.csv"], "kind": "file", "value": "data/shift_7.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-s1-01074-twin
cadenceworks
ctx_compress-s1-01074
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] overlap_calendar_page returns the upper bound for every value instead of clamping into the range. It lives in src/cadenceworks/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 map...
46,333
2,316
63
124
false
synth:3101026652240151760@1.5709
synth
[{"aliases": ["src/cadenceworks/rules.py", "rules.py"], "kind": "file", "value": "src/cadenceworks/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/shift_7.csv", "shift_7.csv"], "kind": "file", "value": "data/shift_7.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/cadenceworks/tables.py", "tables....
ctx_compress-s1-01075
tariffledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_receipt skips the first page; page 1 returns what should be page 2. It lives in src/tariffledger/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 tariffledger layou...
71,793
3,072
89
100
false
synth:4270343209761367898@2.22812
synth
[{"aliases": ["src/tariffledger/cli.py", "cli.py"], "kind": "file", "value": "src/tariffledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/tariffledger/schema.py", "schema.py"], "kind": "file", "value": "src/tariffledger/schema.py"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["cli.py", "src/tariffledger/cli.py"], ...
ctx_compress-s1-01075-twin
tariffledger
ctx_compress-s1-01075
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_receipt skips the first page; page 1 returns what should be page 2. It lives in src/tariffledger/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 tariffledger layou...
38,046
1,902
55
100
false
synth:4270343209761367898@2.22812
synth
[{"aliases": ["src/tariffledger/cli.py", "cli.py"], "kind": "file", "value": "src/tariffledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/tariffledger/schema.py", "schema.py"], "kind": "file", "value": "src/tariffledger/schema.py"}, {"ali...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/invoice_4.csv", "invoice_4.csv"]...
ctx_compress-s1-01076
principalforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] extend_session raises ValueError when a value itself contains an equals sign. It lives in src/principalforge/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the princip...
56,231
2,811
82
99
false
synth:7410810965830419721@1.3624
synth
[{"aliases": ["src/principalforge/compat.py", "compat.py"], "kind": "file", "value": "src/principalforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/policy_1.csv", "policy_1.csv"], "kind": "file", "value": "data/policy_1.csv"}...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/token_0.csv", "token_0.csv"], "k...
ctx_compress-s1-01076-twin
principalforge
ctx_compress-s1-01076
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_session raises ValueError when a value itself contains an equals sign. It lives in src/principalforge/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the princip...
32,536
1,626
50
99
false
synth:7410810965830419721@1.3624
synth
[{"aliases": ["src/principalforge/compat.py", "compat.py"], "kind": "file", "value": "src/principalforge/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/policy_1.csv", "policy_1.csv"], "kind": "file", "value": "data/policy_1.csv"}...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/policy_1.csv", "policy_1.csv"], ...
ctx_compress-s1-01077
invoiceledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_credit produces doubled dashes for consecutive separators. It lives in src/invoiceledger/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 invoiceledger layo...
51,245
2,562
89
96
false
synth:926103687678853379@1.20528
synth
[{"aliases": ["src/invoiceledger/limits.py", "limits.py"], "kind": "file", "value": "src/invoiceledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/balance_2.csv", "balance_2.csv"], "kind": "file", "value": "data/balance_2.csv"...
[{"answers": ["src/invoiceledger/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-s1-01077-twin
invoiceledger
ctx_compress-s1-01077
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] allocate_credit produces doubled dashes for consecutive separators. It lives in src/invoiceledger/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 invoiceledger layo...
29,115
1,455
49
96
false
synth:926103687678853379@1.20528
synth
[{"aliases": ["src/invoiceledger/limits.py", "limits.py"], "kind": "file", "value": "src/invoiceledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/balance_2.csv", "balance_2.csv"], "kind": "file", "value": "data/balance_2.csv"...
[{"answers": ["src/invoiceledger/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-s1-01078
rosterkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_booking returns the upper bound for every value instead of clamping into the range. It lives in src/rosterkit/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 r...
34,000
1,700
52
65
false
synth:3248209558958985487@1
synth
[{"aliases": ["src/rosterkit/rules.py", "rules.py"], "kind": "file", "value": "src/rosterkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/rosterkit/tables.py", "tables.py"], "kind": "file", "value": "src/rosterkit/tables.py"}, {"ali...
[{"answers": ["src/rosterkit/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01078-twin
rosterkit
ctx_compress-s1-01078
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_booking returns the upper bound for every value instead of clamping into the range. It lives in src/rosterkit/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 r...
28,825
1,441
36
65
false
synth:3248209558958985487@1
synth
[{"aliases": ["src/rosterkit/rules.py", "rules.py"], "kind": "file", "value": "src/rosterkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["src/rosterkit/tables.py", "tables.py"], "kind": "file", "value": "src/rosterkit/tables.py"}, {"ali...
[{"answers": ["src/rosterkit/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-s1-01079
tileflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_fix_all leaves a trailing dash when the title ends in punctuation. It lives in src/tileflow/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 tileflow layout....
108,520
3,072
124
153
false
synth:3401574306447985074@2.3707
synth
[{"aliases": ["src/tileflow/core.py", "core.py"], "kind": "file", "value": "src/tileflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/tile_5.csv", "tile_5.csv"], "kind": "file", "value": "data/tile_5.csv"}, {"aliases": ["data/bearing_7...
[{"answers": ["src/tileflow/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-s1-01079-twin
tileflow
ctx_compress-s1-01079
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_fix_all leaves a trailing dash when the title ends in punctuation. It lives in src/tileflow/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 tileflow layout....
71,595
3,072
86
153
false
synth:3401574306447985074@2.3707
synth
[{"aliases": ["src/tileflow/core.py", "core.py"], "kind": "file", "value": "src/tileflow/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["data/tile_5.csv", "tile_5.csv"], "kind": "file", "value": "data/tile_5.csv"}, {"aliases": ["data/bearing_7...
[{"answers": ["src/tileflow/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-s1-01080
scopeforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_principal_safe grows linearly instead of doubling. It lives in src/scopeforge/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 scopeforge layout. [tool_use kind...
72,334
3,072
106
123
false
synth:3061260271835514198@1.32612
synth
[{"aliases": ["src/scopeforge/rules.py", "rules.py"], "kind": "file", "value": "src/scopeforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ...
[{"answers": ["src/scopeforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01080-twin
scopeforge
ctx_compress-s1-01080
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_principal_safe grows linearly instead of doubling. It lives in src/scopeforge/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 scopeforge layout. [tool_use kind...
37,707
1,885
70
123
false
synth:3061260271835514198@1.32612
synth
[{"aliases": ["src/scopeforge/rules.py", "rules.py"], "kind": "file", "value": "src/scopeforge/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, {"aliases": ...
[{"answers": ["src/scopeforge/rules.py", "rules.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01081
reorderledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_lot grows linearly instead of doubling. It lives in src/reorderledger/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 reorderledger layout. [tool_use kind=s...
56,728
2,836
99
124
false
synth:5582044565962768766@1.81523
synth
[{"aliases": ["src/reorderledger/limits.py", "limits.py"], "kind": "file", "value": "src/reorderledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, ...
[{"answers": ["src/reorderledger/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-s1-01081-twin
reorderledger
ctx_compress-s1-01081
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] reserve_lot grows linearly instead of doubling. It lives in src/reorderledger/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 reorderledger layout. [tool_use kind=s...
38,871
1,943
61
124
false
synth:5582044565962768766@1.81523
synth
[{"aliases": ["src/reorderledger/limits.py", "limits.py"], "kind": "file", "value": "src/reorderledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["docs/changelog.md", "changelog.md"], "kind": "file", "value": "docs/CHANGELOG.md"}, ...
[{"answers": ["src/reorderledger/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-s1-01082
beaconledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 drops override values of 0, False and empty string. It lives in src/beaconledger/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 beaconledger layout. ...
37,568
1,878
77
84
false
synth:1805063710731736982@1
synth
[{"aliases": ["src/beaconledger/cli.py", "cli.py"], "kind": "file", "value": "src/beaconledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/tile_2.csv", "tile_2.csv"], "kind": "file", "value": "data/tile_2.csv"}, {"aliases": ["data/beacon_...
[{"answers": ["src/beaconledger/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-s1-01082-twin
beaconledger
ctx_compress-s1-01082
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 drops override values of 0, False and empty string. It lives in src/beaconledger/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 beaconledger layout. ...
18,689
1,000
43
84
false
synth:1805063710731736982@1
synth
[{"aliases": ["src/beaconledger/cli.py", "cli.py"], "kind": "file", "value": "src/beaconledger/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["data/tile_2.csv", "tile_2.csv"], "kind": "file", "value": "data/tile_2.csv"}, {"aliases": ["data/beacon_...
[{"answers": ["src/beaconledger/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-s1-01083
lotflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_pallet leaves a trailing dash when the title ends in punctuation. It lives in src/lotflow/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 lotflow layout. [...
52,317
2,615
100
125
false
synth:8544988214774365605@1.2818
synth
[{"aliases": ["src/lotflow/limits.py", "limits.py"], "kind": "file", "value": "src/lotflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/lotflow/report.py", "report.py"], "kind": "file", "value": "src/lotflow/report.py"}, {"alias...
[{"answers": ["src/lotflow/limits.py", "limits.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most...
ctx_compress-s1-01083-twin
lotflow
ctx_compress-s1-01083
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] transfer_pallet leaves a trailing dash when the title ends in punctuation. It lives in src/lotflow/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 lotflow layout. [...
34,956
1,747
62
125
false
synth:8544988214774365605@1.2818
synth
[{"aliases": ["src/lotflow/limits.py", "limits.py"], "kind": "file", "value": "src/lotflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/lotflow/report.py", "report.py"], "kind": "file", "value": "src/lotflow/report.py"}, {"alias...
[{"answers": ["src/lotflow/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-s1-01084
tileforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_corridor raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/tileforge/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...
102,285
3,072
116
129
false
synth:187965362040912170@1.94732
synth
[{"aliases": ["src/tileforge/audit.py", "audit.py"], "kind": "file", "value": "src/tileforge/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/zone_2.csv", "zone_2.csv"], "kind": "file", "value": "data/zone_2.csv"}, {"aliases": ["src/ti...
[{"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/tileforge/tables.py", "tables.py"...
ctx_compress-s1-01084-twin
tileforge
ctx_compress-s1-01084
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_corridor raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/tileforge/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...
65,499
3,072
76
129
false
synth:187965362040912170@1.94732
synth
[{"aliases": ["src/tileforge/audit.py", "audit.py"], "kind": "file", "value": "src/tileforge/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/zone_2.csv", "zone_2.csv"], "kind": "file", "value": "data/zone_2.csv"}, {"aliases": ["src/ti...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_compat.py", "tests/test_compat.p...
ctx_compress-s1-01085
revisionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] fold_footnote raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/revisionkit/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 revi...
52,520
2,626
83
104
false
synth:378932737413173604@1.25018
synth
[{"aliases": ["src/revisionkit/rules.py", "rules.py"], "kind": "file", "value": "src/revisionkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/index_0.csv", "index_0.csv"], "kind": "file", "value": "data/index_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": ["test_cli.py", "tests/test_cli.py"], "...
ctx_compress-s1-01085-twin
revisionkit
ctx_compress-s1-01085
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_footnote raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/revisionkit/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 revi...
26,593
1,329
47
104
false
synth:378932737413173604@1.25018
synth
[{"aliases": ["src/revisionkit/rules.py", "rules.py"], "kind": "file", "value": "src/revisionkit/rules.py"}, {"aliases": ["tests/test_rules.py", "test_rules.py"], "kind": "file", "value": "tests/test_rules.py"}, {"aliases": ["data/index_0.csv", "index_0.csv"], "kind": "file", "value": "data/index_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": ["audit.py", "src/revisionkit/audit.py"...
ctx_compress-s1-01086
stemdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_chunk averages one value too many once the window is full. It lives in src/stemdesk/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 stemdesk layout. [tool_use ...
23,318
1,165
41
50
false
synth:510343475897799622@1
synth
[{"aliases": ["src/stemdesk/store.py", "store.py"], "kind": "file", "value": "src/stemdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/cue_1.csv", "cue_1.csv"], "kind": "file", "value": "data/cue_1.csv"}, {"aliases": ["src/stemdes...
[{"answers": ["src/stemdesk/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-s1-01086-twin
stemdesk
ctx_compress-s1-01086
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] dedupe_chunk averages one value too many once the window is full. It lives in src/stemdesk/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 stemdesk layout. [tool_use ...
12,018
1,000
27
50
false
synth:510343475897799622@1
synth
[{"aliases": ["src/stemdesk/store.py", "store.py"], "kind": "file", "value": "src/stemdesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["data/cue_1.csv", "cue_1.csv"], "kind": "file", "value": "data/cue_1.csv"}, {"aliases": ["src/stemdes...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["caption_0.csv", "data/caption_0.csv"]...
ctx_compress-s1-01087
waypointforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] interpolate_fix raises KeyError for a missing key instead of returning the default. It lives in src/waypointforge/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 wa...
97,230
3,072
115
126
false
synth:3042781373346333003@1.72457
synth
[{"aliases": ["src/waypointforge/format.py", "format.py"], "kind": "file", "value": "src/waypointforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/beacon_7.csv", "beacon_7.csv"], "kind": "file", "value": "data/beacon_7.csv"}, ...
[{"answers": ["src/waypointforge/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent'...
ctx_compress-s1-01087-twin
waypointforge
ctx_compress-s1-01087
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_fix raises KeyError for a missing key instead of returning the default. It lives in src/waypointforge/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 wa...
59,493
2,974
67
126
false
synth:3042781373346333003@1.72457
synth
[{"aliases": ["src/waypointforge/format.py", "format.py"], "kind": "file", "value": "src/waypointforge/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/beacon_7.csv", "beacon_7.csv"], "kind": "file", "value": "data/beacon_7.csv"}, ...
[{"answers": ["src/waypointforge/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'...
ctx_compress-s1-01088
tiledesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_tile_page raises BeaconDriftError for bytes input in strict mode instead of decoding it. It lives in src/tiledesk/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 mappi...
89,573
3,072
118
129
false
synth:3199052309294383212@2.324
synth
[{"aliases": ["src/tiledesk/compat.py", "compat.py"], "kind": "file", "value": "src/tiledesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/beacon_8.csv", "beacon_8.csv"], "kind": "file", "value": "data/beacon_8.csv"}, {"aliases"...
[{"answers": ["src/tiledesk/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01088-twin
tiledesk
ctx_compress-s1-01088
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] simplify_tile_page raises BeaconDriftError for bytes input in strict mode instead of decoding it. It lives in src/tiledesk/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 mappi...
51,979
2,598
69
129
false
synth:3199052309294383212@2.324
synth
[{"aliases": ["src/tiledesk/compat.py", "compat.py"], "kind": "file", "value": "src/tiledesk/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/beacon_8.csv", "beacon_8.csv"], "kind": "file", "value": "data/beacon_8.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/zone_4.csv", "zone_4.csv"], "kin...
ctx_compress-s1-01089
calendardesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] rotate_slot raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/calendardesk/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 calen...
24,016
1,200
41
46
false
synth:7069159361521517195@1
synth
[{"aliases": ["src/calendardesk/store.py", "store.py"], "kind": "file", "value": "src/calendardesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/calendardesk/report.py", "report.py"], "kind": "file", "value": "src/calendardesk/report...
[{"answers": ["src/calendardesk/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 m...
ctx_compress-s1-01089-twin
calendardesk
ctx_compress-s1-01089
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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 raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/calendardesk/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 calen...
8,944
1,000
21
46
false
synth:7069159361521517195@1
synth
[{"aliases": ["src/calendardesk/store.py", "store.py"], "kind": "file", "value": "src/calendardesk/store.py"}, {"aliases": ["tests/test_store.py", "test_store.py"], "kind": "file", "value": "tests/test_store.py"}, {"aliases": ["src/calendardesk/report.py", "report.py"], "kind": "file", "value": "src/calendardesk/report...
[{"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/quota_2.csv", "quota_2.csv"], "k...
ctx_compress-s1-01090
sectionflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_margin double counts carried amounts when the period closes on a weekend. It lives in src/sectionflow/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 section...
28,173
1,408
63
78
false
synth:6961843878363480563@1
synth
[{"aliases": ["src/sectionflow/limits.py", "limits.py"], "kind": "file", "value": "src/sectionflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/sectionflow/audit.py", "audit.py"], "kind": "file", "value": "src/sectionflow/audit....
[{"answers": ["src/sectionflow/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-s1-01090-twin
sectionflow
ctx_compress-s1-01090
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_margin double counts carried amounts when the period closes on a weekend. It lives in src/sectionflow/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 section...
17,260
1,000
43
78
false
synth:6961843878363480563@1
synth
[{"aliases": ["src/sectionflow/limits.py", "limits.py"], "kind": "file", "value": "src/sectionflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/sectionflow/audit.py", "audit.py"], "kind": "file", "value": "src/sectionflow/audit....
[{"answers": ["src/sectionflow/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-s1-01091
warehousedesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_lot raises WarehouseRangeError for bytes input in strict mode instead of decoding it. It lives in src/warehousedesk/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 mapp...
28,045
1,402
50
63
false
synth:3783544972492728680@1
synth
[{"aliases": ["src/warehousedesk/report.py", "report.py"], "kind": "file", "value": "src/warehousedesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/warehousedesk/compat.py", "compat.py"], "kind": "file", "value": "src/warehoused...
[{"answers": ["src/warehousedesk/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent'...
ctx_compress-s1-01091-twin
warehousedesk
ctx_compress-s1-01091
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] restock_lot raises WarehouseRangeError for bytes input in strict mode instead of decoding it. It lives in src/warehousedesk/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 mapp...
19,813
1,000
34
63
false
synth:3783544972492728680@1
synth
[{"aliases": ["src/warehousedesk/report.py", "report.py"], "kind": "file", "value": "src/warehousedesk/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/warehousedesk/compat.py", "compat.py"], "kind": "file", "value": "src/warehoused...
[{"answers": ["src/warehousedesk/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent'...
ctx_compress-s1-01092
revisionkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] anchor_footnote_page ignores the cap, so late attempts wait for minutes. It lives in src/revisionkit/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 revisionkit layout. [to...
24,260
1,213
39
46
false
synth:1149729045234190829@1
synth
[{"aliases": ["src/revisionkit/io.py", "io.py"], "kind": "file", "value": "src/revisionkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/clause_0.csv", "clause_0.csv"], "kind": "file", "value": "data/clause_0.csv"}, {"aliases": ["data/margin_3....
[{"answers": ["src/revisionkit/io.py", "io.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most rec...
ctx_compress-s1-01092-twin
revisionkit
ctx_compress-s1-01092
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_footnote_page ignores the cap, so late attempts wait for minutes. It lives in src/revisionkit/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 revisionkit layout. [to...
11,387
1,000
21
46
false
synth:1149729045234190829@1
synth
[{"aliases": ["src/revisionkit/io.py", "io.py"], "kind": "file", "value": "src/revisionkit/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["data/clause_0.csv", "clause_0.csv"], "kind": "file", "value": "data/clause_0.csv"}, {"aliases": ["data/margin_3....
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["data/revision_2.csv", "revision_2.csv...
ctx_compress-s1-01093
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] accrue_balance raises InvoiceDriftError for bytes input in strict mode instead of decoding it. It lives in src/receiptdesk/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...
44,343
2,217
70
85
false
synth:1751694932448619444@1
synth
[{"aliases": ["src/receiptdesk/audit.py", "audit.py"], "kind": "file", "value": "src/receiptdesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/tariff_0.csv", "tariff_0.csv"], "kind": "file", "value": "data/tariff_0.csv"}, {"aliases"...
[{"answers": ["src/receiptdesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mo...
ctx_compress-s1-01093-twin
receiptdesk
ctx_compress-s1-01093
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] accrue_balance raises InvoiceDriftError for bytes input in strict mode instead of decoding it. It lives in src/receiptdesk/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...
24,260
1,213
44
85
false
synth:1751694932448619444@1
synth
[{"aliases": ["src/receiptdesk/audit.py", "audit.py"], "kind": "file", "value": "src/receiptdesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/tariff_0.csv", "tariff_0.csv"], "kind": "file", "value": "data/tariff_0.csv"}, {"aliases"...
[{"answers": ["src/receiptdesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's mo...
ctx_compress-s1-01094
policydesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_grant returns the upper bound for every value instead of clamping into the range. It lives in src/policydesk/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the po...
34,157
1,707
62
73
false
synth:1556570983916963232@1
synth
[{"aliases": ["src/policydesk/audit.py", "audit.py"], "kind": "file", "value": "src/policydesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policydesk/schema.py", "schema.py"], "kind": "file", "value": "src/policydesk/schema.py"}, {...
[{"answers": ["src/policydesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01094-twin
policydesk
ctx_compress-s1-01094
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_grant returns the upper bound for every value instead of clamping into the range. It lives in src/policydesk/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the po...
18,172
1,000
42
73
false
synth:1556570983916963232@1
synth
[{"aliases": ["src/policydesk/audit.py", "audit.py"], "kind": "file", "value": "src/policydesk/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["src/policydesk/schema.py", "schema.py"], "kind": "file", "value": "src/policydesk/schema.py"}, {...
[{"answers": ["src/policydesk/audit.py", "audit.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's mos...
ctx_compress-s1-01095
bookingdesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/bookingdesk/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 bookingde...
85,624
3,072
112
127
false
synth:2364373207659616464@2.25885
synth
[{"aliases": ["src/bookingdesk/core.py", "core.py"], "kind": "file", "value": "src/bookingdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/bookingdesk/report.py", "report.py"], "kind": "file", "value": "src/bookingdesk/report.py"}, {"a...
[{"answers": ["src/bookingdesk/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-s1-01095-twin
bookingdesk
ctx_compress-s1-01095
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_window keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/bookingdesk/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 bookingde...
52,539
2,626
65
127
false
synth:2364373207659616464@2.25885
synth
[{"aliases": ["src/bookingdesk/core.py", "core.py"], "kind": "file", "value": "src/bookingdesk/core.py"}, {"aliases": ["tests/test_core.py", "test_core.py"], "kind": "file", "value": "tests/test_core.py"}, {"aliases": ["src/bookingdesk/report.py", "report.py"], "kind": "file", "value": "src/bookingdesk/report.py"}, {"a...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/bookingdesk/core.py"],...
ctx_compress-s1-01096
invoiceworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] settle_tariff returns one item too few on every page. It lives in src/invoiceworks/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the invoiceworks layout. [tool_use ki...
93,419
3,072
117
136
false
synth:5411711098801212768@2.08685
synth
[{"aliases": ["src/invoiceworks/compat.py", "compat.py"], "kind": "file", "value": "src/invoiceworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/tariff_2.csv", "tariff_2.csv"], "kind": "file", "value": "data/tariff_2.csv"}, {"...
[{"answers": ["src/invoiceworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01096-twin
invoiceworks
ctx_compress-s1-01096
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_tariff returns one item too few on every page. It lives in src/invoiceworks/compat.py; `python -m pytest tests/test_compat.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping the invoiceworks layout. [tool_use ki...
45,572
2,278
63
136
false
synth:5411711098801212768@2.08685
synth
[{"aliases": ["src/invoiceworks/compat.py", "compat.py"], "kind": "file", "value": "src/invoiceworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/tariff_2.csv", "tariff_2.csv"], "kind": "file", "value": "data/tariff_2.csv"}, {"...
[{"answers": ["src/invoiceworks/compat.py", "compat.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01097
scopekit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_session_batch skips the first page; page 1 returns what should be page 2. It lives in src/scopekit/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 scopekit layout....
21,165
1,058
51
62
false
synth:219405610819793933@1
synth
[{"aliases": ["src/scopekit/cli.py", "cli.py"], "kind": "file", "value": "src/scopekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/scopekit/audit.py", "audit.py"], "kind": "file", "value": "src/scopekit/audit.py"}, {"aliases": ["src/scopek...
[{"answers": ["src/scopekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-01097-twin
scopekit
ctx_compress-s1-01097
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] narrow_session_batch skips the first page; page 1 returns what should be page 2. It lives in src/scopekit/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 scopekit layout....
9,306
1,000
33
62
false
synth:219405610819793933@1
synth
[{"aliases": ["src/scopekit/cli.py", "cli.py"], "kind": "file", "value": "src/scopekit/cli.py"}, {"aliases": ["tests/test_cli.py", "test_cli.py"], "kind": "file", "value": "tests/test_cli.py"}, {"aliases": ["src/scopekit/audit.py", "audit.py"], "kind": "file", "value": "src/scopekit/audit.py"}, {"aliases": ["src/scopek...
[{"answers": ["src/scopekit/cli.py", "cli.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most rece...
ctx_compress-s1-01098
sectionworks
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_section_all drops override values of 0, False and empty string. It lives in src/sectionworks/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 sectionworks l...
72,199
3,072
89
112
false
synth:5713166426526710547@2.11298
synth
[{"aliases": ["src/sectionworks/compat.py", "compat.py"], "kind": "file", "value": "src/sectionworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/draft_2.csv", "draft_2.csv"], "kind": "file", "value": "data/draft_2.csv"}, {"ali...
[{"answers": ["src/sectionworks/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-s1-01098-twin
sectionworks
ctx_compress-s1-01098
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_section_all drops override values of 0, False and empty string. It lives in src/sectionworks/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 sectionworks l...
44,724
2,236
57
112
false
synth:5713166426526710547@2.11298
synth
[{"aliases": ["src/sectionworks/compat.py", "compat.py"], "kind": "file", "value": "src/sectionworks/compat.py"}, {"aliases": ["tests/test_compat.py", "test_compat.py"], "kind": "file", "value": "tests/test_compat.py"}, {"aliases": ["data/draft_2.csv", "draft_2.csv"], "kind": "file", "value": "data/draft_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": ["compat.py", "src/sectionworks/compat....
ctx_compress-s1-01099
snippetforge
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_lexeme skips the first page; page 1 returns what should be page 2. It lives in src/snippetforge/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 snippetforg...
58,232
2,911
86
107
false
synth:471604049591812906@1.1892
synth
[{"aliases": ["src/snippetforge/report.py", "report.py"], "kind": "file", "value": "src/snippetforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_3.csv", "snippet_3.csv"], "kind": "file", "value": "data/snippet_3.csv"},...
[{"answers": ["src/snippetforge/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's...
ctx_compress-s1-01099-twin
snippetforge
ctx_compress-s1-01099
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_lexeme skips the first page; page 1 returns what should be page 2. It lives in src/snippetforge/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 snippetforg...
31,178
1,558
54
107
false
synth:471604049591812906@1.1892
synth
[{"aliases": ["src/snippetforge/report.py", "report.py"], "kind": "file", "value": "src/snippetforge/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["data/snippet_3.csv", "snippet_3.csv"], "kind": "file", "value": "data/snippet_3.csv"},...
[{"answers": ["src/snippetforge/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's...
ctx_compress-s1-01100
cueflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_batch returns the upper bound for every value instead of clamping into the range. It lives in src/cueflow/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping th...
41,629
2,081
63
78
false
synth:4610388994754156451@1
synth
[{"aliases": ["src/cueflow/audit.py", "audit.py"], "kind": "file", "value": "src/cueflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/caption_4.csv", "caption_4.csv"], "kind": "file", "value": "data/caption_4.csv"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["test_compat.py", "tests/test_compat.p...
ctx_compress-s1-01100-twin
cueflow
ctx_compress-s1-01100
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_batch returns the upper bound for every value instead of clamping into the range. It lives in src/cueflow/audit.py; `python -m pytest tests/test_audit.py -q` shows the failures. Fix it without changing the public signature and add a regression test. [assistant] I will start by mapping th...
26,040
1,302
41
78
false
synth:4610388994754156451@1
synth
[{"aliases": ["src/cueflow/audit.py", "audit.py"], "kind": "file", "value": "src/cueflow/audit.py"}, {"aliases": ["tests/test_audit.py", "test_audit.py"], "kind": "file", "value": "tests/test_audit.py"}, {"aliases": ["data/caption_4.csv", "caption_4.csv"], "kind": "file", "value": "data/caption_4.csv"}, {"aliases": ["s...
[{"answers": ["no"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's most recent test run pass or fail? Answer pass or fail.", "type": "word"}, {"answers": {"file": ["core.py", "src/cueflow/core.py"], "ki...
ctx_compress-s1-01101
snippetflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] tokenise_chunk_once raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/snippetflow/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 ...
85,799
3,072
103
110
false
synth:2422484951502856292@1.57418
synth
[{"aliases": ["src/snippetflow/limits.py", "limits.py"], "kind": "file", "value": "src/snippetflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/stem_5.csv", "stem_5.csv"], "kind": "file", "value": "data/stem_5.csv"}, {"aliases"...
[{"answers": ["src/snippetflow/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-s1-01101-twin
snippetflow
ctx_compress-s1-01101
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their 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_chunk_once raises ZeroDivisionError on the first value and lags by one afterwards. It lives in src/snippetflow/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 ...
46,541
2,327
59
110
false
synth:2422484951502856292@1.57418
synth
[{"aliases": ["src/snippetflow/limits.py", "limits.py"], "kind": "file", "value": "src/snippetflow/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["data/stem_5.csv", "stem_5.csv"], "kind": "file", "value": "data/stem_5.csv"}, {"aliases"...
[{"answers": ["src/snippetflow/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-s1-01102
invoiceflow
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_refund double counts carried amounts when the period closes on a weekend. It lives in src/invoiceflow/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 invoiceflo...
107,890
3,072
111
136
false
synth:5246903153712970623@1.95518
synth
[{"aliases": ["src/invoiceflow/report.py", "report.py"], "kind": "file", "value": "src/invoiceflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/invoiceflow/io.py", "io.py"], "kind": "file", "value": "src/invoiceflow/io.py"}, {"a...
[{"answers": ["src/invoiceflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01102-twin
invoiceflow
ctx_compress-s1-01102
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] net_refund double counts carried amounts when the period closes on a weekend. It lives in src/invoiceflow/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 invoiceflo...
57,655
2,882
67
136
false
synth:5246903153712970623@1.95518
synth
[{"aliases": ["src/invoiceflow/report.py", "report.py"], "kind": "file", "value": "src/invoiceflow/report.py"}, {"aliases": ["tests/test_report.py", "test_report.py"], "kind": "file", "value": "tests/test_report.py"}, {"aliases": ["src/invoiceflow/io.py", "io.py"], "kind": "file", "value": "src/invoiceflow/io.py"}, {"a...
[{"answers": ["src/invoiceflow/report.py", "report.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["pass"], "question": "Did the agent's ...
ctx_compress-s1-01103
stemledger
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_slug misses keys that differ only in case. It lives in src/stemledger/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 stemledger layout. [tool_use kind=se...
41,167
2,058
90
107
false
synth:1521164075370386781@1
synth
[{"aliases": ["src/stemledger/limits.py", "limits.py"], "kind": "file", "value": "src/stemledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/stemledger/io.py", "io.py"], "kind": "file", "value": "src/stemledger/io.py"}, {"alias...
[{"answers": ["src/stemledger/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-s1-01103-twin
stemledger
ctx_compress-s1-01103
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] hyphenate_slug misses keys that differ only in case. It lives in src/stemledger/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 stemledger layout. [tool_use kind=se...
20,029
1,001
52
107
false
synth:1521164075370386781@1
synth
[{"aliases": ["src/stemledger/limits.py", "limits.py"], "kind": "file", "value": "src/stemledger/limits.py"}, {"aliases": ["tests/test_limits.py", "test_limits.py"], "kind": "file", "value": "tests/test_limits.py"}, {"aliases": ["src/stemledger/io.py", "io.py"], "kind": "file", "value": "src/stemledger/io.py"}, {"alias...
[{"answers": ["src/stemledger/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-s1-01104
policydesk
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_nonce ignores the cap, so late attempts wait for minutes. It lives in src/policydesk/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 policydesk layout. [tool_use kin...
80,175
3,072
105
122
false
synth:4983154650106072338@1.59928
synth
[{"aliases": ["src/policydesk/io.py", "io.py"], "kind": "file", "value": "src/policydesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/policydesk/cli.py", "cli.py"], "kind": "file", "value": "src/policydesk/cli.py"}, {"aliases": ["data/token_7.c...
[{"answers": ["src/policydesk/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-s1-01104-twin
policydesk
ctx_compress-s1-01104
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] revoke_nonce ignores the cap, so late attempts wait for minutes. It lives in src/policydesk/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 policydesk layout. [tool_use kin...
55,357
2,767
73
122
false
synth:4983154650106072338@1.59928
synth
[{"aliases": ["src/policydesk/io.py", "io.py"], "kind": "file", "value": "src/policydesk/io.py"}, {"aliases": ["tests/test_io.py", "test_io.py"], "kind": "file", "value": "tests/test_io.py"}, {"aliases": ["src/policydesk/cli.py", "cli.py"], "kind": "file", "value": "src/policydesk/cli.py"}, {"aliases": ["data/token_7.c...
[{"answers": ["src/policydesk/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-s1-01105
headingkit
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_draft keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/headingkit/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 headingkit l...
83,724
3,072
103
110
false
synth:5320134046399555300@2.12292
synth
[{"aliases": ["src/headingkit/format.py", "format.py"], "kind": "file", "value": "src/headingkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/index_7.csv", "index_7.csv"], "kind": "file", "value": "data/index_7.csv"}, {"aliases...
[{"answers": ["src/headingkit/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...
ctx_compress-s1-01105-twin
headingkit
ctx_compress-s1-01105
You compact a coding agent's session so the agent can continue from your summary alone. The session is inside <session> tags: [user] messages, [assistant] text, [tool_use] calls with their kind and file, and [tool_result] outputs. Write the summary in plain markdown with exactly these nine headings, in this order, each...
<session> [user] redact_draft keeps leading spaces in keys so lookups after a semicolon fail. It lives in src/headingkit/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 headingkit l...
53,206
2,660
63
110
false
synth:5320134046399555300@2.12292
synth
[{"aliases": ["src/headingkit/format.py", "format.py"], "kind": "file", "value": "src/headingkit/format.py"}, {"aliases": ["tests/test_format.py", "test_format.py"], "kind": "file", "value": "tests/test_format.py"}, {"aliases": ["data/index_7.csv", "index_7.csv"], "kind": "file", "value": "data/index_7.csv"}, {"aliases...
[{"answers": ["src/headingkit/format.py", "format.py"], "question": "Which file did the agent edit most recently?", "type": "exact"}, {"answers": ["yes"], "question": "Has the agent changed any existing file in the repository yet? Answer yes or no.", "type": "word"}, {"answers": ["fail"], "question": "Did the agent's m...