Submit a skill or command

The catalog is small on purpose: every item is policy-linted and human-reviewed before it ships, and growth is gated on quality — not volume. If you've built a skill or command worth that bar, we'd genuinely like to list it, with your name on it.

The review bar

Every submission runs through the same executable policy lint as our first-party items — these rules are enforced in CI, verbatim:

• ·No dynamic-context shell preprocessing (the !`…` syntax) — anywhere, for any tool.
• ·No network fetches in executable positions (curl/wget/fetch or raw URLs inside code blocks).
• ·Path-safe artifacts only: repo-relative, no absolute paths, no traversal.
• ·No credential-shaped strings — every artifact is scanned by the same redaction pipeline that guards analysis uploads.
• ·Skills must be spec-valid Agent Skills: skills/<name>/SKILL.md with frontmatter name matching the folder.
• ·A fresh review date — items fail CI when their last review goes stale.

Then a human reviews it — for correctness, for whether the guidance is honest, and for whether it earns a slot in a deliberately small catalog. Acceptance lands it in the catalog with author attribution and a link back to you. There is no open publishing and no pay-to-list.

How to submit

Open a submission issue on our public repo with your item's content (the SKILL.md or rule body), which score gap it addresses, and a line on why it beats what's already listed:

We review in the open; expect honest feedback rather than a fast yes. Items that duplicate an existing listing without improving on it are declined — the catalog being small is the point.