agentsclimarketplace

Skillstat

Skill no7z/skillstat

Audit your Claude Code skills: see which actually fire, which are dead weight eating context tokens, and slim them down. Local, zero-dependency, AI-free.

Install
npx -y skills add no7z/skillstat

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 26 days oldThe repository was created 26 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Audit which installed Claude Code, Codex, and Cursor skills actually fire, compare usage across agents, estimate Claude context cost, and safely slim unused Claude user skills. Use when the user asks which skills they use, which agent used them, whether they have too many skills, or what is bloating context. Runs the local skillstat CLI offline.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.8 KB, 610 tokens by cl100k_base, as published. Nobody here has run it

skillstat — audit your installed skills

skillstat is a local, zero-dependency CLI that parses Claude Code, Codex, and Cursor transcripts to report which installed skills actually trigger. Claude provides explicit/auto evidence, Codex provides explicit/observed evidence, and Cursor provides explicit evidence only. Everything runs locally.

How to run it

Prefer a globally installed binary; fall back to npx:

skillstat <command>        # if installed: npm i -g skillstat
npx -y skillstat <command> # otherwise

Choosing the command

Pick the sub-command from what the user is asking:

User intentCommand
"Which skills do I actually use?" / usage breakdownskillstat scan
"Which Codex skills did I use?"skillstat scan --agent codex
"Compare Claude and Cursor"skillstat scan --agent claude,cursor
"Show every skill, even unused ones"skillstat scan --all
"What's filling my context?" / "are my skills expensive?"skillstat cost
"Give me a report I can look at / share"skillstat report -o skillstat-report.html
"Clean up / disable / archive unused skills"skillstat slim (see safety note)
"Undo that / bring my skills back"skillstat slim --restore

Useful flags: --days <n> sets the idle threshold (default 30); --json gives machine-readable output for scan and cost; --all includes never-triggered skills in scan; --agent claude,codex,cursor selects sources.

Workflow

  1. Run skillstat scan (or cost) and summarize the result for the user: how many skills are installed, how many are actually active, and — for cost — the estimated per-session token overhead and how much is wasted on idle skills. Note that context cost is Claude-only, Cursor uses transcript file mtime, and unsupported auto-trigger signals are intentionally omitted.
  2. If they want to act on it, run skillstat cost to quantify the savings before suggesting slim.

Safety note for slim

skillstat slim moves idle user skills (never plugin skills) into ~/.claude/skills-disabled/. It is reversible via skillstat slim --restore.

  • Do not pass -y/--yes on the user's behalf without explicit confirmation — let the user see the list and confirm the move.
  • Always tell the user the move is reversible and how to undo it (skillstat slim --restore).

What ships with it: 16 files

58.6 KB alongside SKILL.md, 10 of them executable

src/

test/

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.