What does this skill do, and when should you use it?
OpenSkills is a universal skills loader for AI coding agents, distributed as an npm CLI (npm i -g openskills). It brings Anthropic's SKILL.md skills system to Claude Code, Cursor, Windsurf, Aider, Codex, and anything that can read AGENTS.md, generating the exact same <available_skills> XML that Claude Code uses. Skills load on demand via progressive disclosure to keep agent context clean, and can be installed from GitHub, local paths, or private Git repos. The repo ships an example skill, my-first-skill, demonstrating the standard SKILL.md format and authoring conventions. Caveats: the GitHub license field shows NOASSERTION while the README claims Apache 2.0, and the project is explicitly not affiliated with Anthropic.
Installs skills with npx openskills install <source> from GitHub repos (e.g. anthropics/skills), local paths, or private Git repos — defaulting to project-local ./.claude/skills, or ./.agent/skills with --universal, or ~/.claude/skills with --global. Syncs a manifest: npx openskills sync writes the <available_skills> XML block into AGENTS.md. Loads skills: npx openskills read <skill-name> is executed by the agent in its shell to pull in full instructions, with the base directory reported for resolving bundled references/, scripts/, and assets/ resources. Additional subcommands (list, update, remove, manage) maintain installed skills.
- A Cursor or Windsurf user who wants to reuse skills from the official Anthropic marketplace (anthropics/skills) without running Claude Code
- A multi-agent user running Claude Code alongside other agents, using --universal mode to install into .agent/skills and avoid directory conflicts
- A team versioning project-specific skills in-repo, installing them via local paths or private Git repositories
- A developer learning the Agent Skills spec, using npx openskills read skill-creator after installing anthropics/skills as an authoring guide
- A heavy agent user who wants instructions loaded on demand to keep the context window focused
How do you install this skill?
- This is a teaching example that produces no real task output; do not invoke it as a productivity skill.
- Skill-level license metadata is NOASSERTION; license coverage exists only at the repository level — verify before use.
- Publisher is not registry-verified; review skill content yourself before installing (the repo's SECURITY.md recommends the same).
- The example references scripts/helper.sh, which does not exist in the example directory — distinguish illustrative text from actual files.
- Local filesystem
Requires Node.js 20.6+ and Git. Prefer zero-install via npx: npx openskills install anthropics/skills && npx openskills sync. Optional global install: npm i -g openskills. Default install target is project-local ./.claude/skills; add --universal for ./.agent/skills in multi-agent setups; add --global for ~/.claude/skills.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/numman-ali/openskills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/examples/my-first-skill" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
1) Install a skill: npx openskills install <github-repo | local-path | private-git-url>. 2) Run npx openskills sync to generate or update the skills manifest in AGENTS.md. 3) When a task matches a skill, the agent runs npx openskills read <skill-name> in its shell to load full instructions (comma-separate multiple reads: npx openskills read foo,bar). 4) Maintenance: npx openskills list to view, npx openskills update to refresh, npx openskills remove <name> to delete.
What are this skill's strengths and limitations?
- Exact Claude Code compatibility: same prompt format, folder structure, and marketplace access — zero migration cost
- Universal across agents: works with anything that reads AGENTS.md (Cursor, Windsurf, Aider, Codex, and more)
- Progressive disclosure keeps skills out of context until needed
- Supports local paths, symlinks, and private Git repos — friendly to closed-source and internal teams
- Skills are plain files inside your project, so they version cleanly
- GitHub license field is NOASSERTION, conflicting with the README's Apache 2.0 claim — verify before commercial use
- Not an official Anthropic product and explicitly unaffiliated; upstream spec changes may lag
- Skill loading assumes the agent can execute shell commands; pure API-call scenarios require the caller to supply an execution environment
- No test suite or maturity evidence beyond npm download badges; end-to-end compatibility across agents is not systematically documented
- Repository has no topics, so discoverability metadata is thin
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| OpenSkills Universal Skills Loader this page | 54 · Use with care | ★ 11k | 9mo ago | NOASSERTION |
| Node Modules Inspector — Dependency Audit Skill | 67 · Recommended | ★ 3k | 1mo ago | MIT |
| Skill Creator | 70 · Strongly recommended | ★ 2.2k | 12d ago | MIT |
| Create New Skills (Meta-Skill) | 62 · Recommended | ★ 1.5k | 8mo ago | — |
| Skill Creator | 61 · Recommended | ★ 20k | 14d ago | MIT |
The README positions OpenSkills against Claude Code's native skills system: both use the same <available_skills> XML prompt format and .claude/skills/ storage, but Claude Code invokes via a Skill("name") tool and is locked to itself, while OpenSkills invokes via npx openskills read <name> and works across agents; for distribution, Claude Code uses the Anthropic marketplace while OpenSkills pulls from GitHub. The README also distinguishes it from MCP: MCP targets dynamic tools (requires a server), while skills are static instructions plus resources (plain files) — different problems, not direct competitors.
How did FollowSkills review this skill?
Pure-documentation example skill: no scripts, no network calls, no external effects; least privilege trivially satisfied; repo provides SECURITY.md guidance. Deducted: unverified publisher identity, and the skill's source attribution relies on repository context without independent confirmation.
Instructions are self-consistent and the cited references/skill-format.md exists in-repo; no execution path means no runtime failure surface; static read capped at 10. Deducted: no tests or abnormal-input handling evidence specific to this example skill.
Frontmatter description gives clear trigger conditions (learning skill creation, testing the loader); clear audience; pure Markdown with no environment dependencies. Deducted: no real task output as an example skill, non-fit boundaries undeclared, and limited practical trigger precision.
Well-layered docs with progressive disclosure; repo provides Apache 2.0 license, versioning and maintenance signals. Deducted: skill-level license metadata is NOASSERTION, no changelog, unclear ownership/update path for the skill itself.
Completes its stated goal as a format demonstration and loader test fixture with directly usable content. Deducted: pedagogical rather than task value, limited marginal benefit over reading Anthropic's spec directly, no output verification evidence.
Content is consistent with the in-repo reference file, providing auditable primary material; static review capped at 5. Deducted: no third-party execution evidence, no independent reproduction or cross-source corroboration.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →