Skill Finder

Ask the user (or extract from context) what they need. Capture:

• Goal: What outcome are they trying to achieve?
• Domain: What domain or technology area? (e.g., testing, data analysis, DevOps, security, writing)
• Platform: Which agent platform? (e.g., GitHub Copilot, Claude Code, Codex, any)
• Constraints: Any must-haves or deal-breakers? (e.g., "must work offline", "needs MCP", "Python only")

If the user's message already contains enough detail, skip the interview and proceed.

Step 2 — Fetch the Catalog

Fetch the latest skill catalog from the awesome-agent-skills repo:

URL: https://raw.githubusercontent.com/heilcheng/awesome-agent-skills/main/README.md

Use the fetch_webpage tool to retrieve the raw README.md content. This is the single source of truth for available skills.

Step 3 — Parse the Skill List

From the fetched README, extract every skill entry. Skills are organized in tables under these sections:

• Official Claude Skills (Document Processing) — docx, xlsx, pptx, pdf
• Official OpenAI Codex Skills — repo/user/admin/system scoped
• Official HuggingFace Skills — dataset, evaluation, training, publishing
• Community Skills — organized by sub-category:
  • Skill Collections (aggregator repos)
  • Document Processing
  • Development & Code Tools
  • Data & Analysis
  • Integration & Automation
  • Collaboration & Project Management
  • Security & Systems
  • Advanced & Research

For each skill, capture:

• Name (with link if available)
• Description (the short summary from the table)
• Category (the section/sub-section it belongs to)
• Source (repo link if available)

Step 4 — Score Similarity

For each skill, assess how well it matches the user's requirements across these dimensions:

Assign each dimension a score from 0–10, then compute a weighted total (0–10 scale).

Step 5 — Generate the Report

Produce a markdown report following the template in report_template.md.

Key rules:

• Rank skills by total similarity score, descending.
• Include all skills scoring 5.0 or higher in the "Top Matches" section.
• Include skills scoring 3.0–4.9 in a "Partial Matches" section (brief table, no deep analysis).
• Omit skills scoring below 3.0 entirely.
• If no skills score 5.0+, say so explicitly and suggest the user create a custom skill.
• For each top match, include a 2–3 sentence rationale explaining the score.
• If a Skill Collection (aggregator repo) is relevant, flag it separately — it may contain sub-skills not individually listed.

Step 6 — Recommend Next Steps

After the report, suggest concrete next steps:

1. Install: For the top 1–2 matches, provide the install command or link.
2. Combine: If no single skill covers everything, suggest combining 2–3 skills.
3. Create: If the gap is significant, suggest using the create-skill skill to build a custom one.

Notes

• The catalog evolves. Always fetch fresh data — do not rely on cached or memorized skill lists.
• Skill Collections (e.g., anthropics/skills, agentskill.sh) are meta-entries containing many sub-skills. Flag them for deeper exploration if relevant.
• Some skills link to external repos. Include the link so the user can inspect the SKILL.md directly.