Agentic Awesome Skills 🧩 Skill Open source

What it does

Agentic Awesome Skills (AAS) is a large local catalog of reusable Agent Skills paired with a read-only MCP control plane called AAS Core. The catalog vendors thousands of SKILL.md playbooks gathered and credited from across the ecosystem, spanning development, testing, security, infrastructure, product and more. The idea behind AAS Core is that your coding agent, not the tool, inspects your project, searches the full local catalog, and chooses the exact skills it needs; the MCP then validates that selection in memory and can persist it as a reproducible aas-stack.json manifest and an immutable plan you review before anything changes on disk.

What it does for your agent

• Catalog search and inspection: search_skills, get_skill, list_skill_files, read_skill_file to browse and read the complete local catalog as inert text
• Composition and review: compose_stack, inspect_stack, diff_stack validate the agent's chosen skill IDs without installing them; export_selection_evidence records why each was chosen
• CLI lifecycle: aas stack validate and aas stack plan produce a per-target plan; an audit command reports command, network, credential and filesystem signals in a selected skill before you install
• Distribution paths: direct npx installs, specialized plugin bundles for Claude Code and Codex, and role-based bundles

Works with

Codex and Claude via the local MCP; direct installs and plugins also target Cursor, Gemini CLI, Antigravity, Kiro, OpenCode and Copilot. The MCP and validation are read-only; planning writes only the plan artifact.

How to install

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

Re-run without --dry-run after review. For Codex or Claude, configure the AAS Core MCP and let the agent select skills, then validate and plan.

Maintenance and safety

MIT-licensed and updated frequently. Installing copies files but does not run them; an agent may act on a skill later. The full catalog includes critical and authorized-use-only offensive skills, and the CLI prints a risk summary and offers a per-skill audit. Apply and recovery are experimental and opt-in.

Who should use it

Teams who want an agent-chosen, reviewable, reproducible skill set rather than a hand-picked handful. Be aware it is a broad aggregation with no editorial curation by design, and the presentation is heavy on marketing, so lean on the audit and dry-run steps.