Scaffolds missing structure files (README, CHANGELOG, tests, skill.yml) from SKILL.md
Scans for OPSEC violations (secrets, personal info, hardcoded paths)
Reviews via dual process — agent checklist + user approval
Stages in a private repo for user review on any device
Publishes by erasing history (force push) and flipping repo to public, then ClawhHub

When to Use

"Release {skill-name} to ClawhHub"
"Is {skill-name} ready for release?"
"Which skills are ready to publish?"
"Prepare {skill-name} for release"

Key Design Decisions

History erased at release — orphan branch + force push gives a single clean commit
One repo per skill — starts private for staging, flipped to public at release
User approval is a hard gate — nothing goes public without explicit "approve"
Boilerplate is auto-generated — only SKILL.md content needs to be written by hand
Security scan verified — uses browser to check VirusTotal + OpenClaw scan results on ClawhHub before declaring release complete

Prerequisites

gh CLI authenticated
clawhub CLI installed
Shell access for git operations
Browser tool available (for ClawhHub security scan verification)

License

MIT

Future Work

Multi-Agent Architecture

When release volume justifies it, refactor from single-agent to multi-agent:

Dispatcher (coordinator)
  ├── Scaffolder (Steps 1-2): boilerplate, validation → flash-lite
  ├── Publisher (Steps 3-10): git ops, staging, OPSEC, publish → flash-lite
  ├── Verifier (Step 11): browser-based scan verification → sonnet
  └── Dispatcher delivers Step 12 summary

Triggers: 5+ skills per session, browser failures cascading into re-runs, rate limit cascades.

Prerequisite: Ship 3-5 more skills through single-agent first. Refactor with evidence, not speculation.

ClawhHub Version Purge

No per-version deletion exists. delete is soft-delete (preserves versions). If a version leaks PII: full delete + re-publish (loses download counts) or contact ClawhHub support. Monitor for clawhub delete-version in future CLI releases.