Skill Workshop: OpenClaw's New Way to Turn Agent Work Into Reusable Skills

OpenClaw has shipped Skill Workshop, a structured workflow for creating and revising reusable skills with a human review step before any skill becomes active.

The Problem

When an agent learns how you do something, you shouldn’t have to explain it again. But skills change future behavior — and unlike a bad answer, a bad skill propagates into every subsequent run.

The naive fix — letting the agent write skills directly — makes skill creation a loaded weapon. One misworded instruction becomes part of how work gets done, invisibly, going forward.

How Skill Workshop Works

When an agent creates or revises a skill through Skill Workshop, it starts as a proposal — stored as PROPOSAL.md, not SKILL.md. The draft isn’t active yet.

The workflow:

1. You ask for a reusable workflow
2. The agent produces a proposed skill (not yet live)
3. You review, tweak, or reject it
4. Only when you approve does it become a real skill

Control UI provides two views:

• Board view — full workshop with pending, applied, rejected, and stale proposals, searchable and inspectable
• Today view — one proposal at a time with a simple question: should this become part of your skill set?

“Tweak” is where Workshop earns its name. Generated skills are often almost right — the wording is off, a step is missing, the fallback needs work. You revise it, and the agent incorporates the feedback into the same proposal with history intact.

Skills can also carry supporting files: response templates, example logs, scripts, or templates alongside the main instruction file.

Status

Skill Workshop is live as of June 3, 2026. The feature is described in OpenClaw’s docs under openclaw skills workshop and the Control UI.