Where this fits

You're about to write a new agent skill for your team—maybe a Jira integration or a code review helper. You sketch out the SKILL.md, think about edge cases, and start coding. But wait: have you checked what already exists? The SkillsMP index holds over 1.6 million public SKILL.md files from GitHub, covering Claude Code, Codex, and ChatGPT. Before you spend an afternoon building from scratch, a quick search can reveal a ready-made skill that does 80% of what you need—or at least a solid starting point.

Why agents benefit

• Avoids duplicate work: Agents can proactively search SkillsMP before generating a new skill, preventing the classic "reinvent the wheel" trap.
• Finds hidden gems: The treasure search intentionally ignores star counts to surface well-written but underappreciated skills that top-star results often bury.
• Filters precisely: Search by keyword, category, occupation, or language—so you can narrow down to exactly the niche you need (e.g., "legal contract review" or "SQL query optimization").
• Judges domain maturity: Before committing to a custom skill, agents can gauge whether a domain already has strong existing solutions, informing build-vs-buy decisions.
• Works with zero config: Anonymous access (50 requests/day) means agents can search immediately without API keys, making it easy to integrate into any workflow.

Practical scenarios

Scenario 1: Pre-build sanity check

Your agent is tasked with creating a skill for summarizing PDF research papers. Instead of jumping straight to code, it runs search.mjs "pdf summarization" --category research. It finds 15 existing skills, one with a clever chunking strategy you hadn't considered. You adapt that idea, saving hours of trial and error.

Scenario 2: Cross-tool discovery

You mainly use Claude Code but wonder if Codex or ChatGPT skills have a better approach for a recurring task. Searching without filtering by platform reveals a skill written for Codex that implements a superior error-handling pattern. You port the concept to Claude Code, improving your skill's robustness.

Scenario 3: Treasure hunt for niche needs

A user asks for a skill that generates haiku about server logs. A normal search returns nothing useful. The agent runs treasure.mjs "haiku"—which deliberately skips star-sorted results—and finds a quirky, well-documented skill from a solo developer with 2 stars. It's exactly what you needed, and you just saved an afternoon of creative coding.

Add it to your agent workflow

Install the skill directly via npx:

npx skills add yan-labs/yan-skills --skill skillsmp

Then use it from your agent. A basic search looks like:

node scripts/search.mjs "code review" --limit 10

For deeper exploration, try the treasure hunt:

node scripts/treasure.mjs "code review" --category development

Both commands support --json for easy parsing into your agent's decision loop. Optional API keys (500 req/day) go in a .env file—never committed.

Compare before adopting

Not all search results are equal. When evaluating a skill you find, check:

• Quality signals: Read the SKILL.md itself—is it well-structured, with clear instructions and examples? Ignore star counts (they reflect the repo, not the skill).
• Maintenance freshness: Check the last commit date. A skill untouched for two years may rely on outdated APIs.
• Alternatives: Run multiple searches with different keywords and filters. The first hit isn't always the best.
• Workflow fit: Does it integrate cleanly with your existing toolchain? A skill for Codex might not port directly to Claude Code without tweaks.

Why it is worth tracking

SkillsMP is the largest public index of agent skills, and its growth mirrors the explosion of agent development. With 1.6M+ entries and active maintenance, it's become a canonical reference for what's already been built. The project's open-source nature (16 stars, but early) and the clever treasure feature signal a thoughtful approach to discovery. Track this skill if you regularly build or recommend agent skills—it will save you from reinventing wheels and help you find hidden gems. For a broader view of agent skills, see OpenAgentSkill for more context and related tools.