Resource
ClawHub skill publishing checklist for clean launches.
A publishable skill should be small, auditable, package-tested, explicit about credentials, and honest about writes, network calls, and rate limits.
Checklist
- Keep SKILL.md concise and move bulky docs into references only when needed.
- Document credentials and never ship secrets or local paths.
- Run syntax checks, extracted package tests, and no-key error smoke checks.
- State external writes, dry-run behavior, rate limits, cache/log paths, and approval expectations.
- Treat ClawHub as a supply-chain surface, not just a gallery.
Operator note
A checklist for publishing OpenClaw skills with clear credentials, package tests, safety notes, and supply-chain hygiene.
This resource exists to make AI work visible, bounded, and supportable: scoped workflows, clear approvals, artifacts, logs, and rollback before autonomy.
When to use it
Use this checklist before uploading a skill to ClawHub or sharing it outside a private workspace. Skills are executable supply-chain artifacts, not just markdown prompts.
A good public skill has a small audit surface, clear credential behavior, deterministic scripts, readable errors, and honest documentation about network calls, disk writes, rate limits, and external side effects.
The standard is simple: if a stranger extracted the package and inspected it, they should understand what it can touch, what it cannot touch, and how to test it without risking secrets or live writes.