AuditHub Skills

This repository contains skills for AI agents (e.g., Codex/Claude code) useful for setting up and using AuditHub tools.

Repository layout

  • Each skill lives in its own folder at the repo root.
  • Each skill folder must include:
    • README.md (purpose + author)
    • SKILL.md
    • Optional assets/, references/, or helpers

For multiple related skills, create a single directory with a single README.md, and a sub-directory for each skill.

Installing skills manually

If your agent supports installing skills directly from a repository, prefer that flow first.

If you are using codex locally, symlink a skill folder into your local ~/.codex/skills directory so Codex can load it.

If you are using Claude Code, symlink a skill folder into your local Claude skills directory so Claude Code can load it. For many setups, that directory is ~/.claude/skills.

Example (from repo root):

mkdir -p ~/.codex/skills
ln -s "$(pwd)/" ~/.codex/skills/audithub-skills

Claude Code example (watch out, claude code does not recurse in the skill directory):

mkdir -p ~/.claude/skills
ln -s "$(pwd)/<skill-dir>" ~/.claude/skills/<skill-dir>

Contribution guidelines

  • Keep changes minimal and DRY.
  • Every skill must have SKILL.md and README.md. The README.md must include the purpose and author.
  • If you add templates, place them in assets/ and keep them small.
  • If you add references, place them in references/ and keep them scoped.
  • Prefer ASCII in files unless the skill requires Unicode.
  • Add or update tests/examples when behavior changes.

Skills

  • orca: Set of skills for setting up OrCa fuzzing campaigns and analyzing OrCa results.
  • defi-vanguard: Set of skills for setting up DeFi Vanguard scans and analyzing their results.