agentsclimarketplace

Audit

Skill brillcp/Laconic/skills/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

Install
npx -y skills add brillcp/Laconic --skill audit

Assembled 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

  1. Read ~/.claude/settings.json → list enabledPlugins (only true entries).
  2. For each enabled plugin, read ~/.claude/plugins/installed_plugins.json to find its installPath.
  3. From installPath, read plugin.json for description; enumerate: skills (skills/*/SKILL.md), agents (agents/*.md), commands (commands/*.md), hooks (entries in plugin.json), MCP servers (mcpServers in plugin.json).
  4. 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 SessionStart output 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_tokens per plugin per turn. Compute total across all enabled plugins.
  5. 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 ~tokens descending — biggest savings first.
  6. Ask user which to disable. Never edit settings unprompted.
  7. On confirmation, edit ~/.claude/settings.json to set enabledPlugins.<name> = false. Do not uninstall — disable is reversible.

Rules

  • Never disable laconic itself.
  • Never disable plugins the user just installed (<24h).
  • Never touch extraKnownMarketplaces or installed_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)"

Keep looking

Skills are one crate of 608,666. 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.