sickn33/agentic-awesome-skills

AAS Core is a local, agent-first control plane for AI coding agents to discover, select, and validate skills from a catalog of agentic capabilities.

The tool addresses the problem of giving coding agents access to a complete, curated skill library while maintaining human control and reproducibility. Rather than ranking or recommending skills, AAS Core lets the agent inspect the full project surface—architecture, behavior, data, integrations, testing, security, deployment—then search and compose its own skill stack. The agent's selection is validated in memory through a read-only compose_stack tool, and can be persisted as an immutable aas-stack.json manifest alongside selection evidence that records the agent's reasoning. The CLI provides validation and planning commands to preview changes before applying them, and a browser-based Workbench allows offline review of stack and plan files without filesystem access.

Developers should adopt this tool if they want their coding agents to have agency in skill selection while maintaining auditability and control. It suits projects where reproducibility and review of agent decisions matter—the persisted manifest and evidence sidecar let teams understand exactly which skills were chosen and why. The tool is designed as a local control plane, not a hosted service; the catalog and Workbench are companion discovery surfaces, but the canonical repository and local MCP are where the actual composition and validation happen. Direct installs, plugins, bundles, and workflows remain available as distribution and compatibility layers around the core.

Maintainers typically respond to new issues and pull requests within hours. Work in the issue tracker is dominated by bug and enhancement labels.