1999azzar/chaining-mcp/skill-finder

Skill Finder & Manager

Unified tool for discovering, evaluating, installing, and authoring Agent Skills.

Supported Workflows

1. Find & Discover: Search the open agent skills ecosystem (skills.sh, GitHub registries) and recommend verified skills.
2. Install: Download and configure skills from official curated collections, experimental indexes, or custom Git repositories.
3. Scaffold & Create: Generate valid, production-ready new skills with required frontmatter, reference docs, and helper scripts.
4. List & Audit: Inspect currently installed skills, check versions, and review lockfile status.

1. Finding and Evaluating Skills

Use when a user asks "how do I do X", "find a skill for X", or seeks capabilities for a specific domain (testing, design, deployment, database).

Discovery Steps

1. Search CLI:

``bash npx skills find "<query>" ``

1. Browse Repositories:

Check well-known skill sources (openai/skills, vercel-labs/agent-skills, anthropics/skills).

1. Quality Verification:

• Prefer skills with proven install counts (1,000+ installs).
• Check GitHub star counts and maintenance activity.
• Inspect SKILL.md for clear tool boundaries and valid YAML frontmatter.

Presentation Format

When recommending a skill, output:

• Name and primary function
• Source repository and install count
• Installation command
• URL or reference link

2. Installing Skills

Local Installation Commands

Use the bundled helper scripts in scripts/:

• List Curated Skills:

``bash python3 scripts/list-skills.py python3 scripts/list-skills.py --format json ``

• List Experimental Skills:

``bash python3 scripts/list-skills.py --path skills/.experimental ``

• Install from GitHub:

``bash python3 scripts/install-skill-from-github.py --repo <owner>/<repo> --path <path/to/skill> python3 scripts/install-skill-from-github.py --url https://github.com/<owner>/<repo>/tree/<branch>/<path> ``

• Install via Skills CLI:

``bash npx skills add <package> ``

Installation targets $CODEX_HOME/skills or ~/.agents/skills.

3. Creating and Scaffolding New Skills

When creating a new skill, follow standard Agent Skill architecture:

Required Directory Structure

<skill-name>/
├── SKILL.md              # Required: frontmatter and instructions
├── scripts/              # Optional: deterministic executable helpers
├── references/           # Optional: deep reference documentation
└── tests/                # Optional: verification test cases

Frontmatter Schema

---
name: skill-name
description: A concise description of the skill's purpose, triggers, and capabilities. Keep under 100 words.
disable-model-invocation: false # Set true only for manual/slash-command tools
---

Content Guidelines

• Concise trigger conditions: State when the skill should be invoked and when it should not be used.
• Workflow steps: Numbered, sequential procedures with expected inputs and outputs.
• Human prose: Direct technical instructions without promotional phrasing.
• Executable tools over text: Prefer small, testable scripts in scripts/ for deterministic tasks rather than asking the LLM to invent code on the fly.