What problem does it solve?
Choosing and combining AI Agent Skills for a multi-step goal is hard: registries are incomplete, Skills overlap or conflict, and installing unvetted third-party Skills carries security risks. This Skill turns a natural-language goal into the smallest compatible, audited Skill Stack with controlled installation.
Core Features & Use Cases
- Dynamic workflow derivation: Decomposes the user's end goal backward into steps with clear inputs, outputs, access boundaries, and success conditions, instead of forcing a fixed template.
- Multi-source discovery and ranking: Searches the local Skill index, registries (skills.sh, agentskill.sh), GitHub, and OpenCLI through four query lenses, then ranks candidates with hard safety gates and a weighted rubric including community adoption.
- Audited, consent-based installation: Provides staged dry-run installation with file fingerprints, conflict analysis, project Skill Stack profiles, SVG recommendation cards, and a post-install recall check.
- Use Case: A user says "help me run a weekly content workflow." The Skill derives the actual steps, finds a primary Skill plus a fact-check helper, verifies both pass safety checks, and installs them into a project-local profile only after approval.
Quick Start
Use the agent-skill-stack skill to turn my goal into a minimal, compatible, project-specific Skill Stack and explain the recommendation in plain language.
