SkillKit — Package Manager for AI Agent Skills
Install, translate, and share skills across 46 AI coding agents like Claude Code, Cursor, Codex, Copilot.
- Source repo
- rohitg00/skillkit
- Stars
- ★ 1.6k
- Last updated
- 4mo ago
- License
- Apache-2.0
- Primary language
- TypeScript
- FA score
- 40/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformChatGPT · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer who wants to install Anthropic's official skills into Claude Code with
skillkit add anthropics/skills - Not a fit if
- Users who cannot install a Node.js/npm toolchain
- Environments sensitive to install size that reject the default 418-package install
- Source review
- 40/100 · Major gaps 3 safety controls not found
What does this agent do, and when should you use it?
SkillKit is a package manager for AI agent skills that solves the problem of incompatible skill formats across different agents. It installs skills from 31 sources including GitHub, GitLab, local paths, and gists, and automatically translates between formats for 46 agents. The core CLI offers commands like add, translate, sync, and recommend, with optional packages for an interactive TUI, REST API server, P2P mesh network, and agent messaging. It also includes session memory compression, AI skill generation, team manifest sync, security scanning, and codebase analysis. The project is Apache 2.0 licensed, has 757 tests, and is CI-validated.
SkillKit reads and converts skill files like SKILL.md, installing them into agent-specific directories such as .claude/skills/ or .cursor/skills/. It clones skill repositories from GitHub, GitLab, local paths, or gists via skillkit add, performing a security scan. skillkit translate converts skills to other agent formats. skillkit sync deploys skills to agent config files. skillkit recommend analyzes the project stack and suggests relevant skills. It also provides skillkit serve to start a REST API on port 3737, an MCP server, and a Python client. skillkit memory compress extracts session learnings, and skillkit generate creates new skills using multi-source context.
- A developer who wants to install Anthropic's official skills into Claude Code with
skillkit add anthropics/skills - A team that needs to share the same skill set across Cursor, Codex, Copilot, and other agents, using translate and sync commands
- A developer who wants automatic skill recommendations based on their project stack, using
skillkit recommend - A team that wants to automate skill updates in CI/CD pipelines, using
skillkit cicd init - A user who wants to save a webpage as a skill from Chrome, using the extension and importing the generated SKILL.md
- A team that needs to distribute agents across multiple machines, using
skillkit mesh initto create an encrypted P2P network
How do you install or deploy this agent?
Install globally via npm, pnpm, or bun:
npm install -g skillkit
# or
pnpm add -g skillkit
# or
bun add -g skillkitAlternatively, use npx skillkit without installation. Optional packages: npm i -g @skillkit/tui, @skillkit/api, @skillkit/mesh, @skillkit/messaging.
How do you use this agent?
Initialize and install skills:
skillkit init
skillkit add anthropics/skillsTranslate a skill:
skillkit translate my-skill --to cursorSync to agents:
skillkit syncStart the REST server:
skillkit serve
curl "http://localhost:3737/search?q=react+performance"What are this agent's strengths and limitations?
- Supports 46 agents, one skill source works everywhere
- Installs from 31 sources covering 400K+ skills
- Automatic format translation saves duplicate effort
- Advanced features: security scan, recommendations, session memory
- Slim install mode reduces dependencies to 118 packages
- Requires Node.js environment and npm toolchain
- Full install pulls 418 packages, potentially heavy for some users
- AI generation features depend on external API keys or local models
- Skill source quality varies, relies on community curation
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| SkillKit — Package Manager for AI Agent Skills This agent | 40 · Major gaps | CLIFree + model costs | ★ 1.6k | 4mo ago | TypeScript | ChatGPT · Claude Code |
| RepoBrain – Repo Intelligence Engine | 49 · Major gaps | Agent plugin / skillFree + model costs | ★ 1.4k | 3d ago | Python | Codex · Claude Code |
| LeanCTX Context Intelligence | 75 · Good | CLIFree + model costs | ★ 3.9k | today | Rust | Codex · Claude Code · OpenAI API · Claude API |
| NTM (Named Tmux Manager) | 73 · Some gaps | CLIFree + model costs | ★ 459 | 1d ago | Go | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: installation requires user confirmation (interactive selection), security scan mentioned (42/42 pass), but no least-privilege principle stated, no data flow transparency, no sensitive data handling, dependency security only via npm audit claim (0 vulns), external effects include writing to multiple agent config dirs, no rollback mechanism, source attribution clear (official and community sources listed). Deductions: missing data flow transparency, sensitive data handling, rollback; least privilege only partially shown.
Evidence: README and tests consistent, tests cover multiple adapters, CI runs multi-version tests, but dependency availability not verified (no lockfile or dependency list), failure messages only mention CLI catches missing optional packages and prints hint, but no detailed error handling. Deductions: dependency availability unproven, failure message details insufficient.
Evidence: clear target users (AI coding agents), rich scenarios (install, translate, recommend, etc.), supports 46 agents, environment fit (Node>=18, npm/pnpm/bun), but capability boundaries not explicit (e.g., security scan limitations), trigger precision (commands) clear but no auto-trigger conditions. Deductions: capability boundaries and trigger precision partially insufficient.
Evidence: clear information architecture (README with nav), detailed install notes (multiple methods), naming stable (skillkit and sk), rich examples, license clear (Apache-2.0), version present (v1.24.0), but no known limitations, no changelog, maintenance responsibility unclear (no maintainer or contributing guide). Deductions: missing known limitations, changelog, maintenance responsibility.
Evidence: high output usability (clear command output), high marginal value (solves multi-format problem), but cost-benefit not quantified (no performance benchmarks or resource usage). Deductions: cost-benefit evidence insufficient.
Evidence: README claims (e.g., 46 agents, 400K skills) lack traceable sources, tests provide partial verification, but facts and inferences not separated. Deductions: claim traceability insufficient, facts and inferences not separated.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- The tool writes to multiple agent config directories, potentially modifying user project files; use with caution.
- Security scan is static only and does not guarantee skills are free of malicious code.
- Dependency security relies solely on npm audit claims without lockfile verification.
- No rollback mechanism provided; installation may be hard to undo.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Headline: Native Hermes Agent support [Hermes Agent](https://github.com/NousResearch/hermes-agent) by Nous Research landed this February and crossed 100K stars in weeks. It ships with first-class skills at ~/.hermes/skills/ and self-evolving skill generation. Until now, using SkillKit with Hermes meant hacking the Claude adapter or writing skills twice. v1.24.0 ends that. Hermes is a first-class target alongside Claude Code, Cursor, and Codex. What you get …View on GitHub ↗ -
v1.23.0 v1.23.0 — Slim install, safer picker, brand tags
Release notes excerpt
Highlights - Slim install path. Four power features (TUI, REST server, peer mesh, agent messaging) are now optional packages. Default install is full-featured; npm install -g skillkit --omit=optional cuts to the core for a 75% smaller footprint. - Safer first-time agent picker. No more accidental writes to 32 adapter dirs. Default option is the detected agent. All supported agents moved to the bottom with a confirm gate. Multi-select pre-ticks only the detected agent. …View on GitHub ↗ -
Release notes excerpt
What's New Complete --json and spinner coverage (#88, #89) - 47 commands now support --json for structured CI/CD output - 29 commands have loading spinners for I/O operations - Added: create, enable/disable, eval, sync, tap, translate, save, read - Spinners: audit, fix, test, plugin, team, methodology, hook, remove Bug fix - save --json --agent now correctly installs to agents (was skipping file copy in JSON mode)View on GitHub ↗
FAQ
Is SkillKit completely free?
How are permissions handled when installing skills?
Is there any loss during skill translation?
How do I uninstall installed skills?
skillkit remove <skill> or skillkit remove --source org/repo for bulk removal, or --all to remove everything.