AI Agent Skills is an independent discovery directory for reusable capability packages that extend coding agents. It indexes open-source skills for Claude Code, Codex CLI and similar tools, grouping them by task and linking users back to the maintainer's repository for installation and documentation. It is not affiliated with Anthropic or OpenAI.
What the directory offers
Skills are searchable by keyword, category and author. The catalogue covers areas such as code generation, frontend and backend work, databases, testing, DevOps, security, browser automation, design, research, writing and productivity. Detail pages surface the upstream repository, install guidance and available project metadata rather than repackaging the skill as a closed marketplace download.
- A large browsable catalogue of open-source agent skills and templates
- Category, author and keyword discovery
- Links to upstream source code, README files and installation instructions
- Automated quality and safety checks before a listing is marked released
- Statistics for released skills, authors, repository stars and recent additions
- Editorial guides explaining skills, installation and related agent concepts
- A public GitHub index for machine-readable discovery
Catalogue and pricing
The directory is free to browse and does not sell a subscription plan. Its live statistics page reported more than 12,000 released skills and over 1,500 authors at the time of review, with the index refreshed hourly. Those totals are a snapshot and can change as repositories are added, updated or removed.
GitHub star totals shown by the directory are repository popularity signals, not installation counts. A listed skill may also depend on a paid API, external model or commercial service, so using the directory itself for free does not mean every indexed workflow has no operating cost.
Selection and security
AI Agent Skills says listings are collected from public repositories and skill indexes, then pass automated quality and safety checks before release. Its editorial standard favours installable projects with readable documentation and recent activity. Maintainers continue to own and update the upstream source.
Directory inclusion is not a security certification. Skills can contain scripts, request broad file access, call external services or instruct an agent to perform consequential actions. The site explicitly advises users to inspect permissions, scripts and network access before installing. Teams should pin trusted versions where practical, review changes and test unfamiliar skills in a restricted environment.
Our assessment
AI Agent Skills is useful for developers who know the outcome they want but not whether someone has already packaged the necessary agent instructions and tooling. Its category structure is easier to explore than searching GitHub from scratch, while the upstream links preserve access to the actual source and maintainer history.
The value of any individual listing depends on its documentation, maintenance and fit with the target agent. Similar names can hide very different permissions or quality. Treat the directory as a research starting point: inspect the repository, understand every command and dependency, verify compatibility with the current agent version and adopt only the skills that meet the organisation's security and maintenance standards.