Agent skills · every one read in full
Agent skills, audited before you install them
14,816 SKILL.md skills for Claude Code, Cursor, Codex and other agents. We read each skill’s source and rate it SAFE, CAUTION or BLOCK, with the findings to show why: 9,826 SAFE, 4,250 CAUTION, 740 BLOCK.
Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
Cite
Cite this page
ModelsAtlas. “Agent skills”. Retrieved 30 Oct 2026. https://modelsatlas.com/skills
Skills by agent
Where the skill’s own text says it works
Claude Code2,413Claude Desktop206Cursor1,353OpenClaw1,114Codex1,079GitHub Copilot451Gemini CLI321Windsurf232
10,939 skills name no agent in their text and are not listed here.
Skills by job
Curated jobs, not raw GitHub tags
Coding & review2,696Review diffs, enforce conventions, write and refactor code in a language or framework.Testing339Write and run tests, test-driven development, acceptance and end-to-end checks.Security506Threat modelling, penetration testing, secret scanning and vulnerability triage.Docs & writing606READMEs, changelogs, specs, articles and slide decks.Data & analysis596SQL, databases, spreadsheets, notebooks and charts.DevOps & cloud674Containers, CI/CD, deployment, monitoring and cloud platforms.Design & front-end606UI review, design systems, CSS and accessibility checks.Research544Literature search, papers, citations and structured investigation.Automation1,843Scripts, browser and office chores, scheduled and repeatable multi-step jobs.
What the audits found
Across all 14,816 audited skills
- Critical or high finding1,227
- Network use observed941
- Shell commands observed338
- File reads or writes observed603
- Hard-coded secret found198
Highest trust
All SAFE skills by score →| # | Skill | What it does | Verdict | Score | Stars |
|---|---|---|---|---|---|
| 1 | verification-before-completionobra/superpowers | Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions… | ✓SAFE B | 89 | 307,592 |
| 2 | writing-plansobra/superpowers | Use when you have a spec or requirements for a multi-step task, before touching code | ✓SAFE B | 89 | 307,592 |
| 3 | diagnosing-bugsmattpocock/skills | Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow. | ✓SAFE B | 89 | 305,893 |
| 4 | decision-mappingmattpocock/skills | Ask which skill or flow fits your situation. A router over the skills in this repo. | ✓SAFE B | 89 | 305,019 |
| 5 | continuous-learningaffaan-m/everything-claude-code | [DEPRECATED - use continuous-learning-v2] Legacy v1 stop-hook skill extractor. v2 is a strict superset with instinct-based, project-scoped, hook-reliable learning. Do not invoke v1: when continuous learning, session… | ✓SAFE B | 89 | 291,905 |
| 6 | continuous-learning-v2affaan-m/everything-claude-code | Instinct-based learning system that observes sessions via hooks, creates atomic instincts with confidence scoring, and evolves them into skills/commands/agents. v2.1 adds project-scoped instincts to prevent… | ✓SAFE B | 89 | 291,905 |
| 7 | architecture-diagramnousresearch/hermes-agent | Dark-themed SVG architecture/cloud/infra diagrams as HTML. | ✓SAFE B | 89 | 258,210 |
| 8 | arxivnousresearch/hermes-agent | Search arXiv papers by keyword, author, category, or ID. | ✓SAFE B | 89 | 258,210 |
| 9 | verifyfacebook/react | Use when you want to validate changes before committing, or when you need to check all React contribution requirements. | ✓SAFE B | 89 | 251,875 |
| 10 | testfacebook/react | Use when you need to run tests for React core. Supports source, www, stable, and experimental channels. | ✓SAFE B | 89 | 251,857 |
Recently audited
- verification-before-completion obraSAFE · 29 Oct 2026
- diagnosing-bugs mattpocockSAFE · 29 Oct 2026
- continuous-learning affaan-mSAFE · 29 Oct 2026
- architecture-diagram nousresearchSAFE · 29 Oct 2026
- qa garrytanSAFE · 29 Oct 2026
- caveman-discover juliusbrusseeSAFE · 29 Oct 2026
Recently blocked
- build-zoom-contact-center-app anthropicsBLOCK · Prompt injection
- minimax-xlsx minimax-aiBLOCK · Privilege escalation / persistence
- huggingface-lora-space-builder huggingfaceBLOCK · Prompt injection
- ljg-paper-river lijigangBLOCK · Code injection
- expo-cicd-workflows expoBLOCK · Prompt injection
Questions
What is an agent skill?
A folder with a SKILL.md file that teaches an AI agent how to do one job, and sometimes scripts it can run. Agents such as Claude Code, Codex and Cursor read these files.
How do you decide SAFE, CAUTION or BLOCK?
We read the source at one exact commit in layers: provenance, static code, the instructions the agent is given, and an inventory check. Each finding is weighed by severity. The fifth layer, a sandboxed run, is not yet run for skills, and each page says so. Of 14,816 skills, 9,826 are SAFE, 4,250 CAUTION and 740 BLOCK.
Are blocked skills hidden?
No. All 740 BLOCK pages stay public and show exactly which lines caused the verdict, so you can check a skill someone recommended to you.
Is a popular skill a safe one?
Not by itself. 1,227 of the 14,816 skills have at least one critical or high finding, whatever their star count. Every list here is ranked by trust score first and stars second.
Browse A–Z
Every skill, alphabetically · 100 per page
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- What is an agent skill?
- How do you decide SAFE, CAUTION or BLOCK?
- Are blocked skills hidden?