What problem does it solve?
Working with Agent Skills (SKILL.md folders) requires discovering trustworthy candidates, judging their quality, installing them to the right provider paths, and authoring new ones that follow structural best practices — all of which is error-prone without a guided workflow.
Core Features & Use Cases
- Discovery and evaluation: Search GitHub, skills.sh, and skill registries, then score candidates on trigger quality, workflow, gates, evidence, and risk before recommending.
- Gated installation: Install skills to provider-specific destinations (Claude, Cursor, Codex, and others) with approval gates, conflict handling, and copy/symlink modes.
- Authoring and review: Create or improve skills with a lobby-and-references structure, then enforce quality with the
skill-review.mjs done-gate script.
- Use Case: You want a code-review skill for Claude Code. The skill searches registries, inspects real SKILL.md content, ranks candidates by installs and recency, and installs your pick after you approve the destination.
Quick Start
Ask the agent to find and install a skill for a task, for example: find a highly rated code review skill and install it for Claude Code.