Agent Skills

Personal agent skills maintained for AgentVouch and local agent tools.

This repository keeps authored skills in one predictable layout, validates them before publishing, and records AgentVouch listing metadata in registry.json.

Layout

skills/
  <skill-name>/
    SKILL.md
    agents/openai.yaml        # optional UI metadata
    references/               # optional supporting docs
    assets/                   # optional reusable templates/assets
registry.json
scripts/
  validate-skills.mjs
  publish-agentvouch.mjs

Rules

• Keep one skill per skills/<skill-name>/ directory.
• Every skill must have exactly one top-level SKILL.md.
• Skill names use lowercase hyphen-case and should match the directory name.
• Keep repo-level documentation at the repository root, not inside skill folders.
• Preserve upstream attribution for mirrored skills; do not publish third-party work as original wallet-authored work.
• Publish free repo-backed AgentVouch listings by default.
• Authored skills in this repository are MIT licensed; preserve original licenses for any mirrored third-party skills.

Discovering External Skills

Agents can discover task-specific capabilities through the Vercel Agent Skills directory:

npx skills find <query>
npx skills add <owner/repo> --skill <skill-name>

See AGENTS.md for installation guidance, source-review expectations, and rules for bringing third-party skills into this repository.

See SKILLS_ORGANIZATION.md to set up this repository as a shared, Git-managed skill source for Codex, Claude, and Cursor on another machine.

Validate

npm run validate

The validator now includes OKF-style knowledge-base checks: concept frontmatter, reserved index.md / log.md files, internal links, and registry consistency.

Format KB Metadata And Indexes

npm run kb:format

This applies the repo's OKF-inspired conventions to skill/reference frontmatter and regenerates indexes.