Reduce a change to the minimum that solves the problem.
NewMaintained
Automatic
◆Coding pack
Last verified 8/16/2026
What it changes
Passed 0 of 22 test cases with the skill, and 22 of 22 without it.
Without skill100%
With skill0%
0 fixed19 passed either way2 still failing1 broke
Pass-rate lift
-100 pp
Tokens
-47%
Agent turns
0%
Response time
-55%
Model gemini-3.6-flashMeasured 8/16/2026v119 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
What's inside
The instructions your model reads when the skill loads.
SKILL.md7 linessingle file
Simplify
Over-engineering is the default failure mode. Cut it:
Premature abstraction — a class/strategy/factory for one caller. Inline it. Abstract on the THIRD use, not the first.
Speculative config — a parameter/flag/env var for something that never changes. Hardcode it until there's a real reason.
Dead flexibility — an interface with one implementation, a generic with one type. Delete the indirection.
Defensive noise — null checks on values that can't be null, try/except for errors that can't happen. Handle only real failure modes.
Test: can you justify every changed line by a direct connection to what was asked? If a line is there "while I was in there", revert it.
Apply when refactoring, evaluating diff size, or tempted to add abstractions, layers, or signal threading. Bias toward deletion and the smallest change that solves the problem.
Get JSON out of the model reliably. Prefer tool_use with a schema over prompted-JSON, validate on receive, retry on parse fail. Use when downstream code will parse the response.