A Claude Code plugin marketplace. Home of claudit — audit and optimize your Claude Code configuration — plus the repo-level tooling used to author marketplace plugins.
| Plugin | Problem solved |
|---|---|
| claudit | Audit and optimise your Claude Code config. Caches current Claude Code ecosystem knowledge for any subsequent agent task. |
Add quickstop as a plugin marketplace, then install claudit:
/plugin marketplace add acostanzo/quickstop
/plugin install claudit@quickstopOr install from a local clone:
git clone https://github.com/acostanzo/quickstop.git
claude --plugin-dir /path/to/quickstop/plugins/clauditAudit and optimize your Claude Code configuration. Caches current Claude Code ecosystem knowledge so any subsequent agent task — building a skill, configuring an MCP, authoring CLAUDE.md, debugging hooks — can read it via /claudit:knowledge instead of re-fetching docs.
- Recursive CLAUDE.md discovery: root, subdirectory,
.claude/rules/,CLAUDE.local.md, and@importreferences - Automatic scope detection — comprehensive audit inside a git repo, global-only outside
- Research-first: subagents fetch official Anthropic docs before analysis; over-engineering detection is the highest-weighted scoring category
- Decision memory annotates recommendations with past context (team-shared, committable)
- Optional PR delivery with educational inline comments
- Knowledge cache at
~/.cache/claudit/with version-based + 7-day TTL invalidation, exposed to any agent task via/claudit:knowledge [domain|all]
Commands: /claudit, /claudit:knowledge, /claudit:refresh, /claudit:status
Claudit's knowledge cache is general-purpose. Once it's been populated (any /claudit or /claudit:refresh invocation does this), any agent task that needs current Claude Code ecosystem knowledge can read it via /claudit:knowledge instead of re-fetching docs.
| Domain | Covers |
|---|---|
ecosystem |
MCP servers, plugins, hooks, skills, sub-agents |
core-config |
Settings, permissions, CLAUDE.md, memory system |
optimization |
Performance patterns, over-engineering detection |
In a skill or agent prompt, invoke /claudit:knowledge ecosystem (or whichever domain you need) and use the output as expert context. The skill checks freshness and auto-refreshes stale domains transparently. If claudit is not installed, fall back to your own research.
=== CLAUDIT KNOWLEDGE: ecosystem ===
[cached research content]
=== END CLAUDIT KNOWLEDGE ===
Knowledge source: cache (fresh, fetched 2026-03-22) | Domains: ecosystem
Manual refresh: /claudit:refresh [domain|all]. Status: /claudit:status.
Repo-level skills for plugin authors (not distributable plugins — these live in .claude/):
| Command | Purpose |
|---|---|
/smith <name> |
Scaffold a new plugin with correct structure and conventions |
/hone <name> |
Audit an existing plugin's quality (8-category scoring with interactive fixes) |
Both tools dispatch research agents to fetch the latest Anthropic plugin docs before operating, ensuring scaffolds and audits reflect the current spec.
See the Claude Code plugin documentation for plugin authoring and marketplace details.
MIT