Datasheet · A prompt line
No comments
Do not add comments to the code.
Folklore, as cited: a common rule in coding-agent prompts
Hurts on Coding, Haiku 5.5 with thinking off (−3.6 points).
Coding, Haiku 5.5 with thinking off
Pooled across 2 notebooks: −3.6 points, 95% interval −5.5 to −1.7; measured once.
- UNRESOLVEDNotebook #087−2.2 points (−5.8 to +1.0), 432 pairs, cost ×0.85signed verdict 534
- INHIBITSNotebook #088−5.4 points (−9.5 to −1.6), 432 pairs, cost ×0.82signed verdict 540
Where it acts
Its effect sits in Trade profit (−35.9) and Moving averages (−15.4); no other family moved clearly.
- Moving averages−15.4−30.1 to −2.678 tasks
- Trade profit−35.9−52.6 to −18.678 tasks
The other 9 families, none moved clearly
- AMM swap output+9.0−4.5 to +23.178 tasks
- Business days+0.0+0.0 to +0.078 tasks
- Fees in whole cents+0.0+0.0 to +0.079 tasks
- Merging intervals+0.0+0.0 to +0.079 tasks
- Parsing durations+0.0−9.5 to +8.979 tasks
- Run-length encoding+0.0+0.0 to +0.078 tasks
- Comparing versions+0.0+0.0 to +0.079 tasks
- URL slugs+0.0+0.0 to +0.079 tasks
- Top k by score+0.0+0.0 to +0.079 tasks
Read from 864 paired tasks in 2 notebooks, each task against its own notebook’s two controls, with every interval widened for the 11 families read. No notebook preregisters a family: this says where to look, and a notebook says whether it holds. The runs behind it are on each notebook’s page.
Why it hurts
"No comments", on the same 96 tasks, by the way each miss went wrong: wrong fixed-point scaling in pnl 5 with it against 0 without; floored fee before swap 2 with it against 5 without; rounded away fractional seconds 2 with it against 1 without.
OPSIN sorted every miss with the part and without it, on the same tasks of Notebook #088, into the ways of failing it had named for this setting, blind to which agent made each. Descriptive: the verdicts above are the result.
Put this datasheet in a README
The badge shows what the signed record says about this part, in the setting measured most, and links here. It follows the record as new notebooks publish. Add ?suite= and a suite’s id to show another setting.
[](https://opsinlab.io/parts/5ba390692f7369f3)Measure it in a setting it has not been measured in: propose it. As JSON for an agent: /v1/parts/5ba390692f7369f3. Every part: the atlas.