#107107th of 308 in Skills, plugins and prompts
agentskills
Open specification for SKILL.md folders that give AI agents reusable instructions
- Stars
- 26.5k
- License
- Apache-2.0
- Last commit
- Aug 2026
- Language
- Python
14,070 on 27 Mar 202626,545 on 23 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Defines the Agent Skills format: a folder with a SKILL.md file holding name and description metadata plus instructions, and optional scripts, references and assets. Agents load skills in three stages (discovery, activation, execution), so only names and descriptions sit in context until a task matches. Originally developed by Anthropic and released as an open standard; the repo holds the spec and docs, not a runtime.
Who it is for: Developers writing skills or adding skill support to agent clients
Strengths
- Minimal format: one SKILL.md with name and description is enough
- Progressive disclosure keeps context small with many skills installed
- Skills can bundle scripts, references and templates in one version-controlled folder
- Supported by multiple agent clients, listed in a client showcase
Weaknesses
- Specification and docs only; it is not an agent or runtime
- Behavior depends on how each client implements skill loading
- No tagged releases; versioning of the spec is not stated in the README
- Skill quality and security of bundled scripts are left to authors
What it needs
- no GPU
Quick answers
Is agentskills open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Safety checks
Checks passed: 7 of 8 (19 Oct 2026)
Checked commit 69ef37e ↗ (opens in a new tab) · content sha256:dfd73865fb23…
- Pass
No hardcoded keys or tokens
No key matching a known secret format in 138 files
- Pass
No binaries or archives in agent or server folders
No binary or archive in 2 files
- Pass
- Pass
- Fail
A private way to report a vulnerability
No security policy and private vulnerability reporting is off
security page ↗ (opens in a new tab), reporting setting ↗ (opens in a new tab)
- Pass
- Pass
No fake prerequisite or installer steps
No installer or prerequisite step in 2 files
- Pass
No email recipients added since the last checked version, no hidden BCC
No address added since the check on 2026-10-18 (commit 69ef37e) (0 in the code)
N/A: 10 checks do not apply here or had no data
- N/A
Package points back to this repo
Applies to MCP servers only
- N/A
Listed in the official MCP Registry under the owner's name
Applies to MCP servers only
- N/A
Latest release has build provenance
Applies to MCP servers only
- N/A
No advisories at CVSS 7 or higher
No lockfile at the repo root
- N/A
No install-time scripts
Applies to MCP servers only
- N/A
No dangerous CI workflow
No GitHub Actions workflows
- N/A
HTTPS endpoints, no credentials in shipped config
Not a plugin and not an MCP server
- N/A
No override, concealment or secret-file instructions
No agent-facing instruction files
- N/A
No shell commands built from input, no unchecked URL opening
Applies to MCP servers only
- N/A
HTTP endpoints check auth or stay on localhost
Applies to MCP servers only
Facts, not pass or fail
- Bundles scripts
- Yes, 1
- Hooks
- Yes
- A bin folder
- No
- MCP config file
- No
- Calls other MCP tools
- No
- Domains its agent files link to
- None
- Hosts in its code new since the last check
- None
Read on 19 Oct 2026 from the repo's files, npm, PyPI, deps.dev, OSV and the MCP Registry. We ran nothing from the project. N/A means the check does not apply or we could not read its data. Not part of the score: what each check looks at and what it cannot see. A result looks wrong? Tell us on GitHub (opens in a new tab).
Also in Skills, plugins and prompts
See all 308| Rank | Project | Score |
|---|---|---|
| 1 | impeccableDesign skill with 24 slash commands for AI coding agents' frontend output | 96 out of 100 |
| 2 | cavemanCuts token use in coding agents with terse replies and compressed inputs | 95 out of 100 |
| 3 | skillsAgent skills for planning, TDD, debugging and code review | 94 out of 100 |
| #4 | eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses | 94 out of 100 |
| #5 | oh-my-claudecodeMulti-agent orchestration plugin and CLI for Claude Code | 94 out of 100 |