audit-claude-marketplace

Audit Claude Marketplace

A curated collection of Claude Code skills.

Skills

Skill	Description
audit-to-plan	Language-agnostic code audit — discovers all issues and writes a structured FIX_PLAN.md with ISSUE-NNN IDs
quality-loop	Iterative fix-verify-commit workflow — drives the codebase to  clean on a dedicated branch, then opens a PR

Installation

Platform-specific install steps for this marketplace.

Claude Code (primary)

This repo contains a Claude Code manifest at .claude-plugin/plugin.json.

Options:

• Install directly from the Claude Code client using the GitHub repo URL (client discovers .claude-plugin automatically).
• Install from a local path for development:

git clone <repo-url>
# point your Claude Code client at the local repo path

Notes:

• The client will list available skills found under skills/.
• Always run audit workflows with --dry-run first to avoid accidental commits.

Copilot CLI

This repo includes a Copilot CLI manifest at .github/plugin/plugin.json.

Install from GitHub:

gh copilot plugin install <repo-url>

Install from local path (development):

gh copilot plugin install /path/to/local/repo

Verify installation:

gh copilot plugin list

Codex / OpenAI / generic LLM wrappers

Codex and many LLM clients are not plugin systems. Use a small wrapper that:

1. Loads skills/<skill>/SKILL.md to get the instructions.
2. Composes a prompt with user arguments.
3. Sends the prompt to the LLM API and returns the output.

Example (minimal Python wrapper):

from pathlib import Path
import os
from openai import OpenAI

skill = Path('skills/audit-to-plan/SKILL.md').read_text()
user_input = 'run audit --scope diff'
prompt = skill + "\n\nUser: " + user_input

client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])
resp = client.responses.create(model=os.environ.get('PREFERRED_MODEL', 'gpt-4o-mini'), input=prompt)
print(resp.output_text)

Safety:

• Always test with --dry-run or --simulate before enabling any auto-apply behavior.
• Use environment variables for secrets (OPENAI_API_KEY) and never commit them.

See platform_templates/codex/README.md for more examples.

Gemini CLI

Gemini clients vary. If the client supports plugins, install using a client-specific manifest. Otherwise use a wrapper similar to the Codex example above. See platform_templates/gemini/README.md for guidance.

Local usage & developer testing

Clone the repository and run the included helper scripts locally (safe defaults):

git clone <repo-url>
cd market
python3 -m venv .venv
. .venv/bin/activate
python3 -m pip install -U pip
# (optional) pip install openai pytest

# Validate skill manifests
python3 scripts/validate-skills.py

# Convert sample FIX_PLAN.md to JSON
python3 scripts/fix_plan_parser.py --md tests/fixtures/sample_FIX_PLAN.md --json FIX_PLAN.json

# Run the reference runner (dry-run by default)
python3 scripts/quality_loop.py --plan docs/FIX_PLAN.md --dry-run

Suggested workflow (safe)

1. Run audit-to-plan with --scope diff to collect issues.
2. Inspect docs/FIX_PLAN.md (and FIX_PLAN.json if you generated it).
3. Run quality-loop with --dry-run to preview commits.
4. When comfortable, run with --apply and --no-push to create commits locally, then push and open a PR.

---

If you need platform-specific examples added to README (e.g., a step-by-step for a specific Claude Code client or Gemini deployment), tell me which client and I will add exact commands.

Usage

Examples:

• Claude / Copilot style:

  /skill audit-to-plan --scope full --output ISSUES.md
  

• Codex / generic LLM wrapper (pseudo):

  from pathlib import Path
  prompt = Path('skills/audit-to-plan/SKILL.md').read_text()
  prompt += "\n\nUser: run audit --scope diff"
  # send prompt to LLM and print response
  

Platform support

This marketplace includes manifests and guidance for multiple platforms:

• .claude-plugin/plugin.json — Claude Code (primary)
• .github/plugin/plugin.json — Copilot CLI manifest
• platform_templates/codex/ — guidance for Codex / generic LLM wrappers
• platform_templates/gemini/ — guidance for Gemini CLI integrations

To add a platform, create a manifest in platform_templates/<platform>/ and document the install steps there.

Adding a New Skill

1. Create skills/<skill-name>/SKILL.md with required frontmatter: