addlightnesslisted

# addlightness "Simplify, then add lightness." — Colin Chapman, Lotus. This skill owns the full trim-and-benchmark loop: it measures code weight, removes fat without changing behavior, verifies the change passes the equivalence gate (a change-magnitude/structural signal, not a behavioral proof), and benchmarks the result. Lower weight and a statistically significant speedup are the only success conditions. The user's invocation carries the target file path(s) as trailing text (e.g. `/addlightness src/parser.js` or "add lightness to lib/foo.py and lib/bar.py"). Read those paths from the request and treat them as the targets. If no path is given, ask which file(s) to trim — do not guess. ## Scope - **This skill (`/addlightness`)** owns the WHOLE pipeline: measure -> trim -> verify -> benchmark. It both reports and edits. - **`/addlightness-review`** is READ-ONLY. If the user only wants a weight report or a list of fat candidates with no edits, route there instead and stop. - **`/addlightness-bench`** is BENCHMARK-ONLY. If the user already has a before/after pair and only wants the timing comparison, route there instead and stop. When in doubt about whether the user wants edits, ask once. Default to the full pipeline only when the request clearly asks to trim, slim, lighten, or speed up the code. ## Pipeline Run these steps in order, per target file. Paths below use `${CLAUDE_PLUGIN_ROOT}`, which Claude Code expands to this plugin's install directory. 1. **Measure