Skip to content

Agent skills: inspect the instructions before reuse

Discover agent skill repositories and learn how to review their instructions, bundled scripts and compatibility before using them.

All collections

A reusable skill can save you from explaining the same procedure every session. It can also carry assumptions about tools, accounts and file access that do not fit your work. This collection helps you find candidate repositories and inspect what a skill actually asks an agent to do. Treat the repository as material to review before granting it access to a real project.

Separate the format from the label

The Agent Skills format uses a folder containing SKILL.md, with metadata and task instructions; it can also include scripts and references. A repository containing the word skills may instead be a tutorial, prompt collection or unrelated training material. Open the candidate and check the actual folder structure. Then read your agent client documentation to confirm how it discovers skills and which features it supports.

Review the entire execution path

Read the instructions from start to finish, including linked local references. Note each executable script, external domain and installation step. Look for actions that write files, send messages or publish changes. Decide which actions fit the task you want to delegate. An instruction saying a script is safe is only an assertion; inspect the script and its dependencies before allowing it to run.

Try a narrow, reversible example

Choose one sample task, such as reviewing a fictional API response or formatting a small test document. Use copies of the input files and omit production credentials. Observe whether the skill reads only the intended material and produces the requested artifact. Test a missing input too: does it ask for the information, fail clearly or invent an answer? Save the version you evaluated so later edits can be reviewed deliberately.

Keep useful instructions maintainable

Prefer a clear procedure with examples you can verify. Check whether the skill assumes a particular operating system, command-line tool or paid service. A large catalogue may contain a few relevant skills alongside many you will never use; evaluate the specific folder rather than adopting everything. Repository metadata cannot prove format compliance or quality. This collection is a discovery aid, and the current index can miss projects or show older metadata.

From this category in our index

283 non-archived matches. This preview is sorted by stars, not a recommendation or security assessment.

obra/superpowers

An agentic skills framework & software development methodology that works.

affaan-m/ECC

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

JuliusBrussee/caveman

🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.

Search all projects in this category

Common questions

Is an agent skill the same as an MCP server?

They serve different roles. A skill packages instructions and supporting materials for a task. An MCP server exposes capabilities through a protocol to a compatible application. A skill might tell an agent when to use a server, but finding one does not automatically install or configure the other.

Can I reuse a skill in another client?

Check the client support and the skill dependencies first. Portable instructions may still rely on a particular tool name, filesystem path or extension. Review the repository licence before copying or adapting its materials, and rerun your sample task in the destination client.

Go to the source

Editorially reviewed 8 November 2026. Repository summaries come from GitHub; metadata can change between checks. Send a correction.