FAQ

FAQ

Is this a general GitHub API tool?

No. It uses GitHub search and repository signals, but its core job is Agent Skill discovery, scoring, risk scanning, installation governance, and lifecycle management.

Should it run for every task?

No. Small tasks should use built-in capabilities or already-installed skills. Search is reserved for high-value, repeated, latest/best, or capability-gap situations.

Can it guarantee third-party skills are safe?

No. It can make risks visible and reject obvious high-risk candidates, but safety scans are heuristic.

Does it install skills automatically?

Not by default. Unknown third-party skills require review and user approval before installation.

Should awesome lists be installed wholesale?

No. Awesome lists can be useful indexes, but linked projects remain candidates. Install individual skill folders only after review.

Why not automatically install every high-star skill?

Stars are a popularity signal, not a safety or fit guarantee. A high-star repository can still be stale, too broad, missing SKILL.md, poorly scoped, incompatible with the current task, or risky to install. See Why Not Auto Install High Star Skills.

Why keep Codex and Claude Code both in scope?

Both can use local skill folders with SKILL.md. The structure is shared; install paths and direct invocation syntax differ.

What is the most important rule?

Prefer the smallest reliable skill set that solves the user's task.

Why not put all of this in the README?

The README should make the project easy to understand and star quickly. The Wiki is for long-form install notes, scoring rules, risk examples, and governance detail.