How Skill Navigator works. Where listings come from.

A directory for finding agent skills by the task you want an AI coding agent to handle. Skills for Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and other agents are listed side by side, so you can compare what they do, how actively they are maintained and how they are licensed before you install one.

Browsing and search are public and need no account. Skill Navigator is independent and not affiliated with any agent vendor.

An agent skill is a folder with a SKILL.md file: a short name and description the agent reads to decide when the skill applies, followed by instructions, plus optional scripts, templates and reference files. Agents load a skill only when a task calls for it, so a library of skills adds expertise without filling every conversation.

Listings come from public GitHub repositories that contain SKILL.md files. Each skill folder is read and parsed; copies of the same skill found in other repositories are grouped, so a skill is listed once, under its original source: an official publisher first, then the earliest repository.

Skills whose automated safety check fails are kept out of the ranked lists.

Each skill gets one score built from public signals:

• Popularity. GitHub stars of the repository, shared across the skills it holds. Large collections and awesome-lists count for less.
• Adoption. How many other GitHub owners keep a copy of the skill in their own repositories.
• SKILL.md quality. Valid frontmatter, a description that says when to use the skill, substantive instructions, examples, bundled scripts and references.
• Safety. Warnings from the automated check lower the score; a failed check pushes a skill far down the list.

Within one repository each further skill steps down a little, so a single large repository can't fill a list. Trending lists sort by star growth over the last 30 days, relative to the repository's size, instead of the score.

Every listed skill's SKILL.md is scanned for risky patterns. It is a pattern match, not a security review or an audit: a pass means none of the patterns matched, not that the skill is safe. Read a skill's files before you install it, especially anything that runs commands.

• Auto-check passedNo risky patterns found in SKILL.md.
• Auto-check: notesWorth knowing: sudo, a well-known installer piped to a shell, a .env mention or unrestricted Bash access.
• Auto-check: warningsRead carefully: credential paths, data-collection endpoints, prompt-injection phrases, hidden characters, URL shorteners or raw IP links.
• Auto-check failedHigh risk: unknown downloads piped to a shell, decoded or remote code run with eval, secrets sent out, destructive commands.

• GitHub. The skill files themselves, plus repository facts such as stars, licence and latest commit, read from GitHub's API.
• The skill authors. Names, descriptions and instructions come from each skill's own SKILL.md. Skills belong to their authors and keep their own licences.

Spotted a wrong detail, want a listing removed, or have a question about how Skill Navigator works? Email maruf@trackrev.io.

For AI tools and feed readers there's a plain-text llms.txt, an RSS feed and a sitemap.