skill-auditlisted

# skill-audit Point it at a skill; it grades that skill against the seven-gate Floor and hands back a gate-by-gate verdict with one concrete fix per miss. It is a lint, not a verdict on your taste: it catches the structural tells a machine can catch, and it says plainly where a human still has to look. ## The gap it closes Ask a bare model "is this skill any good?" and it will write you a confident, agreeable paragraph that checks nothing. The thing that does not survive being done from memory is the *mechanical* read ��� counting whether the trigger is a real prefix or a pile of keywords, whether each hard-fail has a check behind it or is just a sentence, whether the one example shows a catch or the happy path. This skill runs that read every time, the same way, and reports it. ## Trigger Any message beginning with `audit:` fires this skill. Treat the prefix as a command, not a topic. - `audit: <path>` — grade one skill (a directory with a SKILL.md, or a SKILL.md file). - `audit: <folder>` — grade every skill in a folder of skills. - Append `--json` for machine-readable output, `--quiet` to drop the per-gate fix lines, `--log-row` for a one-line row you can paste into build notes. ## How to run it Run the bundled checker and build your answer from its output: ``` python3 skill_audit.py <path> ``` **Resolve `skill_audit.py` from this file, not from the working directory.** It ships beside this `SKILL.md`, and that folder is where the command above has to run — p