ClawHub

Skill Release Lifecycle

Workflows

A lightweight release-quality workflow for OpenClaw / ClawHub skills. Use when deciding whether a skill is ready to publish, verifying a release, and turning post-release feedback into actionable updates.

Install

openclaw skills install skill-release-lifecycle

Skill Release Lifecycle

A lightweight release-quality workflow for OpenClaw / ClawHub skills.

Install

openclaw skills install skill-release-lifecycle --global

Features

🚦 Release Readiness Gate: decide whether a skill is ready to publish or should stay private.

🔍 Release Verification: check install path, activation, safety, metadata, and public-page status after release.

🔁 Feedback → Iteration Loop: turn feedback into ignore / patch / version bump / reposition / deprecate decisions.

🧩 Prompt for Agent: generate a copy-paste prompt you can give to your agent for further release improvement.

What This Will Not Do

  • It does not write SKILL.md from scratch.
  • It does not run ClawHub publish commands directly.
  • It does not replace clawhub-auto-publish.
  • It does not optimize ClawHub ranking, downloads, stars, or homepage placement.
  • It does not certify a skill as stable or fully validated.

How It Works

1. Release Readiness Gate

Check five things before release: utility, identity, safety, UX, and maintenance.

Hard fail:

  • no recurring use case
  • no anti-scope
  • unsafe default behavior
  • secret exposure
  • destructive behavior without explicit protection

Soft fix:

  • unclear install command
  • unclear activation
  • unclear output format
  • missing feedback path
  • weak version/update explanation

2. Release Verification

Before publishing, check local SKILL.md, title, summary, install command, activation, anti-scope, safety, and release status.

After publishing, verify with ClawHub / OpenClaw evidence where available:

clawhub inspect <slug>
openclaw skills search "<keyword>"
openclaw skills install <slug>
openclaw skills info <slug>
openclaw skills check

Public page check if accessible.

If the public page is blocked, report BLOCKED, not PASS.

3. Feedback → Iteration Loop

Classify feedback as:

  • bug
  • confusing activation
  • missing install path
  • scope too broad
  • missing anti-scope
  • safety concern
  • output mismatch
  • UX wording issue
  • strategic repositioning
  • not worth acting on

Then choose:

  • ignore / log only
  • docs patch
  • minor wording patch
  • version bump
  • major repositioning
  • deprecate / merge
  • keep private

Prompt for Agent

Copy this prompt into your agent:

Run the skill-release-lifecycle gate on <skill-name/slug>.

Inspect the current SKILL.md. For each gate (Utility, Identity, Safety, UX, Maintenance), write Pass / Fail / Unclear with a one-line reason and label your evidence source (direct inspection / reference-derived / historical claim / unverified).

If any hard gate fails, do not publish. If soft gates fail, patch within the same session before publishing.

After patching, run:
- clawhub inspect <slug>
- openclaw skills search "<keyword>"
- openclaw skills install <slug>
- openclaw skills info <slug>

Report exact evidence: commands run, observed output, files changed, version bumped, and unresolved blockers.

Do not claim public page verification if it was blocked.

Related Skills

  • hermes-agent-skill-authoring — SKILL.md authoring, frontmatter, validator rules
  • clawhub-auto-publish — detailed ClawHub publish mechanics and CLI flow
  • waste-audit — reference implementation used in examples

Feedback

Found an issue or have feedback? DM me on X: @BeeGeeEth

More in Workflows