{"slug": "skill-workshop-turn-agent-work-into-reusable-skills", "title": "Skill Workshop: Turn Agent Work Into Reusable Skills", "summary": "OpenClaw launched Skill Workshop, a tool that requires agents to submit proposed reusable procedures as reviewable drafts before they become active skills. The system prevents agents from directly writing skill files by routing all skill creation and revision through a proposal workflow that must be approved by users. The workshop aims to prevent flawed agent-generated procedures from becoming permanent parts of automated workflows.", "body_md": "A useful agent should learn the work you keep giving it.\n\nIf you teach an agent how you do something, you should not have to paste the same explanation forever.\n\nWhen that lesson becomes reusable, OpenClaw should show you the draft before it changes future runs.\n\nThat is [Skill Workshop](https://docs.openclaw.ai/tools/skill-workshop).\n\n[Skills](https://docs.openclaw.ai/tools/skills) are how OpenClaw teaches an agent reusable procedures. A skill might be a short checklist for invoice follow-up, a release routine with validation steps, or a repo health workflow with scripts, templates, and examples beside the main instruction file.\n\nA skill is not just Markdown. It changes future behavior.\n\nThat makes skill creation different from normal file editing. If the agent writes a bad answer, you can ignore one answer. If the agent writes a bad skill, that mistake can become part of how future work is done.\n\nSkill Workshop puts a review step in front of that change.\n\n## Proposal First\n\nWhen an agent creates or revises a skill through Skill Workshop, it starts as a proposal.\n\nThe proposal is not active. It carries the draft instruction, any support files, its review state, and the information OpenClaw needs to apply or reject it cleanly.\n\nWhile it is pending, the file is `PROPOSAL.md`\n\n, not `SKILL.md`\n\n. The agent does not run it yet.\n\nSo the loop feels like normal collaboration:\n\n```\nYou:   Make this weekly inbox routine reusable.\nAgent: Drafted a proposed skill.\nYou:   Add the VIP sender rule and make the dry-run step clearer.\nAgent: Revised the proposal.\nYou:   Use it.\n```\n\nThe agent does not need to create skill files by hand or guess where they belong. It does not need one path for chat, another for the UI, and another for the CLI.\n\nIt uses Skill Workshop.\n\n## Shape It Before You Keep It\n\nThe Control UI treats proposed skills as reviewable work, not hidden files you have to hunt down.\n\nBoard view is the full workshop. You can move through pending, applied, rejected, and stale proposals. You can search, inspect, preview support files, and see what changed.\n\nToday view is the faster pass.\n\nIt gives you the next proposal and asks a concrete question: should this become part of your skill set?\n\nUse it when it is ready. Tweak it when it is close. Skip it when it should not become part of your skill set.\n\nTweak is where the workshop matters most.\n\nGenerated work is often almost right. The wording is off. A step is missing. The skill needs a safer fallback. The support file should be a template instead of a script. Skill Workshop turns those fixes into a revision conversation instead of a dead end.\n\nThe proposal remains the object being edited. The agent can revise it, and you can come back to the same proposal with its history intact.\n\n## Skills Can Bring Files\n\nUseful skills often carry supporting material.\n\nA digest skill may need a response template. A debugging skill may need example logs. A deployment skill may need a smoke-test script. A writing skill may need reference snippets.\n\nSkill Workshop keeps those files with the proposal.\n\nSupport files are allowed under the standard skill folders: `assets`\n\n, `examples`\n\n, `references`\n\n, `scripts`\n\n, and `templates`\n\n.\n\nThey are shown in the UI before you apply them. They are scanned with the proposal. They are written with the skill when you apply it.\n\nThe path rules are deliberately narrow. No absolute paths. No traversal. No hidden path segments. No writing outside the skill. A skill can be useful without making the write path loose.\n\nThe same proposal flow works from chat, the Control UI, channels, the CLI, and [Gateway](https://docs.openclaw.ai/gateway).\n\nWhen granted access, agents can create, inspect, revise, apply, or reject proposals through the workshop path. You can review the same proposals from the UI, CLI, chat, channels, or Gateway.\n\nSkill Workshop is the checkpoint before agent work becomes reusable behavior.\n\nThe agent brings back the pattern.\n\nYou shape what stays.", "url": "https://wpnews.pro/news/skill-workshop-turn-agent-work-into-reusable-skills", "canonical_source": "https://openclaw.ai/blog/openclaw-agent-skill-workshop/", "published_at": "2026-06-03 06:00:00+00:00", "updated_at": "2026-06-05 20:17:49.712899+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "ai-products"], "entities": ["OpenClaw", "Skill Workshop"], "alternates": {"html": "https://wpnews.pro/news/skill-workshop-turn-agent-work-into-reusable-skills", "markdown": "https://wpnews.pro/news/skill-workshop-turn-agent-work-into-reusable-skills.md", "text": "https://wpnews.pro/news/skill-workshop-turn-agent-work-into-reusable-skills.txt", "jsonld": "https://wpnews.pro/news/skill-workshop-turn-agent-work-into-reusable-skills.jsonld"}}