<!-- registry-sync: version=15.12.0; skills=2007; stars=44662; updated_at=2026-08-09T06:51:55+00:00 -->

AAS Core — Agentic Awesome Skills

Local, agent-owned skill stacks for coding agents—from complete catalog access to a reproducible, reviewable plan.

Current release: V15.12.0. This release includes AAS Core for complete local catalog search, agent-owned selection, manifest validation, planning, and diagnosis. Apply and recovery remain experimental and outside the supported preview path.

Codex or Claude inspects your project and chooses exact skills from the complete local AAS catalog. AAS Core does not rank or recommend them: its read-only compose_stack tool validates the agent-owned selection in memory, and a client or the aas CLI can persist it as aas-stack.json and produce an immutable plan before any target change.

Read the AAS Core preview guide →

Project
  -> inspected by Codex or Claude (not by AAS)
  -> agent searches and reads the complete local catalog
  -> AAS MCP (local stdio, read-only)
  -> Codex or Claude chooses exact skill IDs
  -> compose_stack validates the selection in memory (read-only)
  -> client or AAS CLI persists aas-stack.json and optional evidence
  -> AAS CLI validate + immutable plan preview
  -> human review (optionally in Workbench)

The reusable SKILL.md playbooks, specialized plugins, bundles, workflows, and direct installers remain important. They are the content, curation, distribution, and compatibility layers around AAS Core—not competing primary products.

This is an independent community project. It is not affiliated with, sponsored by, endorsed by, or authorized by Google. Google, Antigravity, Gemini, and related product names are referenced only to describe compatibility and install targets. The GitHub repository is canonical; the hosted catalog and browser-local Workbench are companion discovery and review surfaces, not a hosted control plane.

AAS Core: Agent-First Preview

The agent composes. You control. AAS keeps the stack reproducible.

AAS Core gives the repository one product model:

• Let the agent choose. The local MCP preserves search_skills, get_skill, compose_stack, inspect_stack, and diff_stack, and adds read-only export_selection_evidence and inspect_selection_evidence; Core does not rank, recommend, exclude, or hide skills.
• Guide capability coverage. MCP session instructions require the agent to evaluate the full project surface—from architecture, domain behavior, data and integrations through testing, security, UX, deployment, and maintenance—then search each applicable capability, compare multiple candidates, cover it with a non-redundant skill or report a catalog gap, and avoid stopping at a minimal shortlist. Core records and validates the resulting selection, but it does not certify semantic completeness.
• Keep the chosen stack and evidence reviewable. A client or the CLI can persist aas-stack.json and the separate aas-selection-evidence.json sidecar in an artifact-dir; the manifest preserves exact agent-selected IDs, while evidence records factual process trace and the agent-declared capability ledger.
• Validate and preview through the CLI. aas stack validate checks the proposal, while aas stack plan produces an immutable, per-target plan without applying it.
• Review in Workbench. The hosted Workbench imports and reviews stack/plan JSON in browser memory; it does not access your filesystem or install anything.
• Retain every useful distribution path. Direct installs, plugins, bundles, workflows, and the full catalog remain available as payload and compatibility surfaces.

[!IMPORTANT] Structural and identity validity does not certify semantic fit, compatibility, setup correctness, operational safety, or safety to apply.

| Surface | Current status | | --- | --- | | Published package | Current npm release; AAS Core status is agent-first-preview | | Catalog search and inspection | Supported preview; local and read-only | | Agent-owned composition | Supported preview; Core validates IDs and structure, not semantic suitability; manifests have a technical maximum of 128 skills | | Stack validation and plan preview | Supported preview; no target skill changes | | Workbench | Browser-local review of stack and plan artifacts | | Selection evidence | Exported and inspected through MCP/CLI contracts; not yet reviewed in Workbench | | Apply and recovery | Experimental, explicit opt-in, outside the supported safety claim | | Semantic suitability certification | Not provided |

Read the AAS Core guide for the exact trust boundaries, current preview status, Codex/Claude setup model, and CLI lifecycle.

Why This Repo

• Agent-first, locally controlled: Codex or Claude inspects the project and chooses from the complete local catalog without uploading your repository to AAS.
• Complete and inspectable: every catalog skill is searchable, readable, and available for agent selection; Core does not certify suitability, compatibility, or operational safety, and metadata is informational rather than an eligibility gate.
• Approval before writes: the durable artifacts are an approved stack and immutable plan, not an opaque one-shot install.
• Installable, not just inspirational: use the compatible legacy installer or plugin distributions when direct delivery is the right path.
• Built for major agent workflows: Claude Code, Cursor, Codex CLI, Autohand Code, Gemini CLI, Antigravity, Kiro, OpenCode, Copilot, and more.
• Broad coverage with real utility: 2,007+ skills across development, testing, security, infrastructure, product, and marketing.
• Inspect before installing: the hosted Skill Workbench reviews agent-produced stack manifests and immutable plans without browser-side installation.
• Focused delivery remains available: specialized plugins package proven sets for web, security, data, docs, DevOps, QA, OSS, or agent/MCP workflows.
• Useful whether you want breadth or curation: install the full catalog, choose a specialized plugin, start with bundles, or compare alternatives before installing.

Why not just search the skills directory?

Direct file search can find candidate prose, but it leaves the result in the conversation. AAS Core adds verified catalog identity, explicit target binding, durable desired state, optional selection evidence, deterministic validation, immutable planning, and dedicated review surfaces. Its value is not choosing better than the coding agent; it is turning the agent's choice into reproducible, inspectable state.

Table of Contents

• AAS Core: Agent-First Preview
• Why This Repo
• Installation
• Recommended Specialized Plugins
• Choose Your Tool
• Quick FAQ
• Bundles & Workflows
• Browse 2,007+ Skills
• Troubleshooting
• Stable Skills Manifest v1
• Support the Project
• Contributing
• Community
• Credits & Sources
• Repo Contributors
• Star History
• License

Installation

For Codex and Claude, start with the AAS Core guide: configure the local MCP, ask the agent to inspect the project and choose exact IDs from the full catalog, review the proposed aas-stack.json, then run CLI validation and planning. The MCP and validation are read-only. Planning writes only the requested plan artifact; it does not materialize skill payloads or AAS managed state in the target.

Use direct installation when your host does not yet have a native AAS Core adapter, when you already know the exact skill IDs, or when you deliberately prefer manual selection:

• Specialized plugins when the job has a clear domain.
• Full library install when you want every skill available in a local skills directory.
• Bundles and workflows when you want role-based recommendations or ordered execution playbooks.

Direct skill install

# Antigravity: preview an exact, agent-selected set before writing.
npx agentic-awesome-skills --antigravity --skills brainstorming,systematic-debugging --dry-run

# Antigravity CLI slash commands (agy): ~/.gemini/antigravity-cli/skills/<skill>/SKILL.md
npx agentic-awesome-skills --agy

The npm installer uses a shallow, release-pinned clone by default so first-run installs stay lighter than a full repository history checkout while matching the published npm package version. Use --tag main only when you intentionally want the current repository tip.

Antigravity watches ~/.agents/skills and may load enough installed instructions to exhaust its context, slow startup, trigger truncation errors, or enter a crash loop. For that target, the installer stops before cloning or writing unless you provide --skills, a metadata filter, or the explicit --all override. The bare npx agentic-awesome-skills command uses the same protected Antigravity target.

The recommended flow is to ask Codex or Claude with the read-only AAS Core MCP configured to inspect the project, search the complete catalog, and choose exact skill IDs. AAS MCP selects and validates IDs but does not install them; the agent or user then previews the direct installation with the command above and repeats it without --dry-run after review.

Other direct-install targets retain the legacy-compatible full-catalog behavior when no selectors are supplied. The CLI prints the catalog's risk summary first: a full install includes critical and authorized-use-only offensive instructions. Installation copies files; it does not execute their commands, but an agent may act on an installe