skills
Teach it a new trick.
Skills are drop-in agent capabilities. Copy a command, run it once, and Anthracode picks the skill up automatically. Every skill here is open source and statically reviewed for dangerous patterns — read the code before you install.
/
#SkillStarsInstall
- 1150.1Kanthropics/skillsanthropics/skills
- 272.0K
- 357.3Kaddyosmani/agent-skillsaddyosmani/agent-skills
- 428.1Kscientific-agent-skillsK-Dense-AI/scientific-agent-skills
- 527.2Khyperframesheygen-com/hyperframes
- 623.2Kplanning-with-filesOthmanAdi/planning-with-files
- 717.9Kclaude-skillsalirezarezvani/claude-skills
- 817.3K
- 95.7Ktrailofbits/skillstrailofbits/skills
- 105.3Kantfu/skillsantfu/skills
- 115.3Ksuperpowers-zhjnMetaCode/superpowers-zh
- 122.2Ksupabase/agent-skillssupabase/agent-skills
- 13423
- 14179well-architected-skillsaws-samples/sample-well-architected-skills-and-steering
- 15142sports-skillsmachina-sports/sports-skills
- 1692veniceai/skillsveniceai/skills
- 1771TerminalSkills/skillsTerminalSkills/skills
- 1837healthcare-agentsajhcs/healthcare-agents
- 1929alva-ai/skillsalva-ai/skills
19 skills · ranked by GitHub stars · all open source, statically scanned, pinned to a reviewed commit.
Want the full registry?
The catalog above is the set we've reviewed. There are thousands more in the open registry — browse them, but read the source yourself before installing anything not listed here.
browse skills.shfaq
How skills work, how we vet them, and how to publish your own.
What is a skill?
A skill is a self-contained set of instructions (and sometimes scripts) that teaches the agent a repeatable capability — a coding convention, a workflow, a tool integration. Anthracode reads installed skills automatically and applies the relevant one when the task matches.How do I install one?
Copy the command on the row and run it in your project. It fetches the skill from its open-source repo into your skills directory. Anthracode discovers it on the next run — no config edit required.How are these vetted?
Every listed skill is open source and goes through a static security review — we read the source for dangerous patterns (network egress, code execution, secret reads, install-time hooks) before it's listed, and pin it to the reviewed commit so upstream can't silently swap in new code. 'Recommended' skills are our highest-trust set. A static review is strong signal, not a guarantee — read the source yourself, exactly as you would any dependency.What do the 'clean' and 'review' tags mean?
'clean' means the static scan found nothing notable. 'review' means it found something worth knowing — usually a benign install script, CI hook, or API-key env read — and the specifics are in the tooltip on the dot. Nothing flagged as dangerous is listed at all.Where do installed skills live?
Locally, alongside your other Anthracode state — ~/.anthracode/ for account-level skills, .anthracode/ in a project for project-scoped ones. Plain files you can read, diff, and git-version.Can I publish my own?
Yes. A skill is just a repo with a skill manifest. Push it to GitHub and anyone can install it with npx skills add <owner>/<repo>. Want it featured here? Open an issue on the Anthracode repo and we'll review it for the catalog.