MCP.Market

labs-graduation skill

by openclaw·openclaw/openclaw·394k stars·MIT

Move a feature out of Control UI Labs, either on by default or still off by default. Use when asked to promote, graduate or ship a Labs feature.

A100/100content scan

Is the labs-graduation skill safe?

Clean: nothing in its files matched our rules. We read 1 file in the folder on 2026-11-14.

No findings.

Install the labs-graduation skill

A skill is a folder. Copy it into your agent's skills folder and the agent loads it when the task matches its description.

git clone --depth 1 https://github.com/openclaw/openclaw.git /tmp/openclaw
mkdir -p ~/.claude/skills
cp -r /tmp/openclaw/.agents/skills/labs-graduation ~/.claude/skills/labs-graduation
available in every project

In the Claude apps, zip the folder and upload it from the Skills settings. The folder on GitHub

The instructions your agent would load

SKILL.md as published, without the frontmatter. Read it on GitHub

Graduate a Labs feature

Graduation removes a feature's Labs switch and makes it a normal feature. Two separate decisions, made by the owner before any code:

  1. Leave Labs: the feature gets a permanent home (its own settings or plugin page) and loses Labs wording.
  2. Default: on by default, or still off by default.

They are independent. #197029 first tried both, then shipped only the first: Advisor left Labs but stayed off. Ask which one is wanted; do not assume "graduate" means "on by default".

Precedents: Swarm on by default (#136514), Tool Search on by default (#154068), native plugin UI on without a Labs opt-in (de50262a1f96), Labs features moved into Settings (#145460), Advisor out of Labs and still off (#197029).

Before the PR

  • Measure cost or risk on the real route if the feature costs money or tokens, or changes behavior for users who never opted in. Today's numbers from mocks or unit tests are not proof. #197074 merged a cache-cost claim from a mock that live providers disproved, and was reverted in #197175. Use the live measurement the feature's own PRs already established, and put the numbers in the PR body.
  • Settle known defects first. Advisor's graduation waited for its fork cache fixes (#197203, #197223, #197224, #197226) because default-on would have charged every user the uncached cost.
  • Find the enable path that remains. After the Labs row goes, users must still be able to turn the feature on from the UI. A bundled plugin uses Settings → Plugins → → Enable; a core feature needs its settings page. Add a switch only if none exists.

The change

  1. Default. Change it only if the owner chose on by default: a bundled plugin's enabledByDefault in openclaw.plugin.json, or the core config default. An existing explicit false (or the feature's opt-out value) must keep working, and an existing true must stay on. Never add a migration that rewrites saved config to change someone's choice.
  2. Labs. Remove the entry from ui/src/pages/labs/labs-registry.ts and everything only it used: labs-page.tsx rows or slots, labs-feature-state.ts entries, strings and tests.
  3. Docs. Move the feature's section out of docs/concepts/experimental-features.md into its own reference page. Leave a one-paragraph note there with the old anchor, saying it graduated, its default, and where to turn it on. Update the "Labs includes ..." list and the "Currently documented flags" table.
  4. Wording. Remove "Experimental" and "OpenClaw Labs" from the manifest description, package description, README, plugin summary, and any user- or model-facing text such as prompt prefixes.
  5. Generated files. Regenerate, never hand-merge: pnpm config:docs:gen, pnpm plugins:inventory:gen. Control UI locale artifacts (.meta.json, .tm.jsonl, catalog-fallbacks.json) stay out of the PR per ui/AGENTS.md; the locale refresh workflow updates them. pnpm ui:i18n:verify must pass.
  6. Upgrade proof. Only for a default change: prove across a real upgrade that no saved entry gets the new default, an explicit off stays off, and an explicit on stays on. A default that stays the same needs no upgrade case; #197029 dropped the one it had built for default-on.

Proof in the PR

  • Control UI screenshots, inspected for private data before upload: Labs before and after, and the enable path working on a real Gateway. Attach them with gh pr comment --attach '#'.
  • The enable path tested end to end: the switch saves the expected config and takes effect without a restart.
  • The feature's tests and the Labs page tests pass; the test-lines ratchet does not grow.
  • Docs claims match code. ClawSweeper caught Advisor's "at most one review per everyMinutes", which was false with everyToolCalls on.

After merge

The new location only appears in builds that include the merge. Tell users the exact path, and that older builds still show the Labs switch.

More skills from openclaw/openclaw

  • A1passwordSet up and use 1Password CLI for sign-in, desktop integration, and reading or injecting secrets.
  • Aacp-routerRoute plain-language requests for Claude Code, Cursor, Copilot, OpenClaw ACP, OpenCode, Gemini CLI, Grok Build, Qwen, Kiro, Kimi, iFlow, Factory Droid, Kilocode, or explicit ACP harness work into either OpenClaw ACP runtime sessions or direct acpx-driven sessions ("telephone game" flow). For coding-agent thread requests, read this skill first, then use only `sessions_spawn` for thread creation. Codex chat binding defaults to the native Codex app-server plugin unless ACP is explicit or background spawn needs ACP.
  • Aadd-model-providerAdd and live-prove a model provider with non-interactive config one-liners, without exposing credentials.
  • Aagent-transcriptRequested GitHub PR/issue agent transcripts: redact, trim, preview, and insert safely.
  • Aapple-notesCreate, view, edit, delete, search, move, or export Apple Notes via the memo CLI on macOS.
  • Aapple-remindersList, add, edit, complete, or delete Apple Reminders and reminder lists via remindctl.
  • Aauto-qaRun OpenClaw-wide autonomous QA and live/stress campaigns across independent subsystem lanes, with verified fixes and a resumable evidence report.
  • AautoreviewStructured code review through the shared OpenClaw agent-skills installation.
  • Abear-notesCreate, search, and manage Bear notes via grizzly CLI.
  • Ablock-kitUse proactively for structured or interactive Slack replies, and when asked to author or validate native Slack Block Kit JSON.
  • AblogwatcherMonitor blogs and RSS/Atom feeds for updates using the blogwatcher CLI.
  • AblucliBluOS CLI (blu) for discovery, playback, grouping, and volume.

All agent skills → · MCP servers