Skip to content
sck_0 edited this page Aug 10, 2026 · 3 revisions

Agentic Awesome Skills Wiki

Welcome to the official wiki for Agentic Awesome Skills (AAS). The older antigravity-awesome-skills name and URL remain compatibility surfaces, but the canonical repository is now sickn33/agentic-awesome-skills.

Last verified: 2026-08-10 against origin/main at bbcd6ef. Current release: v15.12.0, with 2,007 catalog skills.

AAS is an installable library of reusable SKILL.md playbooks for AI coding assistants and agent clients. It covers planning, implementation, debugging, testing, security, infrastructure, product work, documents, data, agent/MCP development, and growth workflows.

Start here

For Codex and Claude, start with AAS Core. The agent inspects the project, searches the complete local catalog, chooses the exact skills, and gives you a reviewable aas-stack.json before any target change.

The short flow is:

project -> agent inspection -> local catalog search -> agent-selected IDs
        -> compose_stack validation -> human review -> CLI validate + plan

Use the direct installer when you already know the skills or your host does not use the AAS Core adapter:

npx agentic-awesome-skills --skills brainstorming,systematic-debugging --dry-run

Use a specialized plugin when the job has a clear domain. Use the full library only when you intentionally want broad plugin-safe coverage.

What you will find

  • A complete, searchable catalog of reusable skills.
  • AAS Core for local, read-only catalog access and agent-owned stack planning.
  • Focused Claude Code, Codex, and portable Agent Plugins distributions.
  • Curated bundles and ordered workflows for common roles.
  • A hosted catalog and browser-local Workbench for discovery and artifact review.
  • Validation, security, provenance, and maintenance documentation.

Useful links

Trust boundaries

The GitHub repository is canonical. The hosted catalog and Workbench are companion discovery and review surfaces, not a hosted control plane.

For AAS Core, the MCP is local stdio, read-only, and offline-capable. It does not install or remove skills, edit host configuration, persist a stack, or apply changes. Skill text remains untrusted content: review the selected skills and their declared risk before activation.

Contributing

Contributions are welcome, including documentation improvements, new skills, corrections to existing playbooks, and stronger validation or security guardrails. Start with the Contributing Guide and the repository's Quality Bar.

License

The project is released under the MIT License.

Clone this wiki locally