sickn33/antigravity-awesome-skills
AAS Core is the local, agent-first control plane for complete catalog discovery, agent-owned selection, stack validation, and planning, backed by 1,987+ agentic skills. Includes CLI, local MCP, catalog, plugins, and Workbench.
What it solves
AI coding agents often struggle to select, validate, and execute complex workflows without leaving behind a messy or unreviewable trail of actions. This project provides a structured way for agents to search a local catalog of specialized skills, compose them into a validated stack, and produce an immutable plan for human review before any changes are applied to a project.
How it works
The system uses a Model Context Protocol (MCP) to allow agents like Claude or Codex to search a local catalog of over 1,900 skills. The agent selects the necessary skill IDs, which are then validated in memory by the AAS Core. This selection can be persisted as a aas-stack.json file, creating a reproducible state. Users can then use a CLI to validate the stack or generate an immutable plan to preview changes before they are executed.
Who it’s for
- AI Coding Agents (such as Claude Code, Cursor, and Codex) that need a standardized way to discover and use specialized capabilities.
- Developers who want to turn opaque, one-shot agent actions into inspectable, reproducible, and versionable execution plans.
- DevOps and QA Engineers looking for specialized, domain-specific skill bundles for security, testing, or infrastructure.
Highlights
- Agent-First Design: Enables agents to perform complete project inspections and select non-redundant skills from a massive local catalog.
- Reproducible State: Converts agent decisions into durable
aas-stack.jsonmanifests and selection evidence. - Safety-Oriented Workflow: Provides an immutable planning phase and a browser-based Workbench for reviewing plans without executing them.
- Extensive Skill Library: Includes nearly 2,000 skills across development, security, data analytics, DevOps, and more.