claudit

claudit

Senior-engineer-grade audits for Claude Code users. Free; telemetry-funded.

Status

v0.1.0 — Scaffold. Plugin monorepo and marketplace manifest are in place. The audit pipeline, finding catalog detectors, report renderer, and telemetry client are not yet implemented. See CHANGELOG.md.

What it is

claudit reads your Claude Code setup — settings, skills, agents, plugins, hooks, MCP servers, and recent session transcripts (with your consent) — and produces an HTML report with:

• A capability band placement: Explorer / Builder / Operator / Architect
• 3–5 personalized "quests" — the highest-leverage changes to make next
• A full deterministic findings list, grouped and collapsible
• Tier-aware language (no jargon for beginners; no condescension for power users)

A second skill, audit-repo, will audit your codebase in a future release.

Install

claudit is a third-party Claude Code plugin shipped from this repo's marketplace manifest. Install is two slash commands run inside an active Claude Code session (not shell commands):

/plugin marketplace add dhabetyes/claudit
/plugin install claudit@claudit

Then run:

/claudit:audit-harness

Why third-party?

claudit ships outside the official Anthropic marketplace by choice. We retain full control over distribution and release cadence; the trade-off is one extra slash command at install time. The plugin runs on your existing Node (Claude Code itself is Node) — no separate runtime, package manager, or daemon.

Telemetry

claudit is free. Telemetry is the business model.

Every audit run posts a UUID-tagged event to a collector we operate, containing:

• Finding types, severity, scores
• Capability band, progression %, quest selections
• Catalog version, audit metadata, approximate token cost

What is never sent: raw transcripts, source files, prompts, code, .env contents, or anything matching common secret patterns. Email is sent only if you submit the report's contact form (and is sent directly from your browser to the collector — the plugin never sees it as part of the audit pipeline).

The full telemetry contract is shown on first run, before the first audit, alongside an estimated token cost. Opt-out is one flag, persistent across runs.

Layout

.claude-plugin/marketplace.json     — third-party marketplace manifest
plugins/claudit/                    — the claudit plugin
  .claude-plugin/plugin.json        — plugin manifest (name, version)
  skills/audit-harness/SKILL.md     — v1 skill: harness audit
  assets/catalog/v1.0.0.yaml        — bundled fallback finding catalog

License

MIT. See LICENSE.