Audit
Audit enabled Claude Code plugins for token cost. Lists each plugin's footprint (tools/skills/hooks/MCP servers), flags candidates to disable, asks user before editing settings. Use when user says "audit plugins", "what plugins cost me tokens", or invokes /laconic audit.From its SKILL.md
npx -y skills add brillcp/Laconic --skill auditAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 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.
SKILL.md
2.7 KB, as published. Nobody here has run it
Goal: help the user reduce per-turn token spend by disabling plugins they don't use.
Steps
- Read
~/.claude/settings.json→ listenabledPlugins(onlytrueentries). - For each enabled plugin, read
~/.claude/plugins/installed_plugins.jsonto find itsinstallPath. - From
installPath, readplugin.jsonfor description; enumerate: skills (skills/*/SKILL.md), agents (agents/*.md), commands (commands/*.md), hooks (entries inplugin.json), MCP servers (mcpServersin plugin.json). - Estimate token cost per plugin. For each loaded-every-turn artifact, sum the bytes that ride in the system prompt and divide by 4 (≈ chars/token):
- skills: each skill's frontmatter
name:+description:only (body loads on invoke — exclude). - agents: full
description:block in frontmatter. - commands: command header +
description:. - hooks: typical
SessionStartoutput bytes (read first line of hook script if shell, else 50 token default). - MCP servers: tool listings — count tools × ~30 tokens (name + short desc).
Sum →
est_tokensper plugin per turn. Compute total across all enabled plugins.
- skills: each skill's frontmatter
- Present a table:
- Plugin · skills · agents · commands · hooks · MCP · ~tokens · recommendation
- Recommendation = "keep" if used recently (check shell history / recent transcripts if available) or core (laconic itself), else "disable".
- Sort by
~tokensdescending — biggest savings first.
- Ask user which to disable. Never edit settings unprompted.
- On confirmation, edit
~/.claude/settings.jsonto setenabledPlugins.<name>=false. Do not uninstall — disable is reversible.
Rules
- Never disable laconic itself.
- Never disable plugins the user just installed (<24h).
- Never touch
extraKnownMarketplacesorinstalled_plugins.json. - Report findings laconic-style: one line per plugin, no preamble.
Output format
plugin skills agents cmds hooks mcp ~tokens rec
caveman 3 0 1 1 0 420 disable (overlaps laconic)
figma 0 0 0 0 1 340 disable (web MCP covers)
gitkraken-hooks 0 0 0 2 0 120 disable (no recent git ops)
laconic 2 0 0 1 0 180 keep
total ~tokens/turn: 1060 → disable 3 → save ~880/turn
Then: "Disable [caveman, figma, gitkraken-hooks]? (y/n)"