ufo-ai/ufo-core/create-skill

Claude Code and agent skill by ufo-ai/ufo-core

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Load when a member asks to create, edit, repair, or save a reusable custom skill for the workspace.

Install

npx vskill@latest install ufo-ai/ufo-core/create-skill

SKILL.md

Create Skill

A saved skill belongs to the workspace. Later turns list it in <available_skills> while the saved set is small, otherwise in the turn's <saved_skills> block; the skill kind's skill_search action finds any saved skill and load_skill loads it. One name is one skill of the workspace: every agent whose use_workspace_skills setting holds loads the same set, and applying a name the workspace already holds edits that skill instead of making a second copy. A saved skill never replaces a built-in skill.

Author

  1. Identify the behavior worth making repeatable and the member phrases that should load it. Check

<available_skills>, <saved_skills>, and the skill kind's skill_search action for overlap. Ask only when a missing choice would change the workflow.

  1. Choose a lowercase slug. Write /workspace/<name>/SKILL.md; its first characters must be ---.
  2. Use exactly this frontmatter shape:

```markdown


name: concise-brief description: "Load when a member asks for a concise project brief with risks and next steps."


```

The name must equal the directory and object name. Make the description a routing trigger of at most 50 words, beginning Load when; name member intent, not the workflow. To narrow the skill to specific agents, add metadata: with agents: [<agent name>, ...] — only those agents' turns list and load it; omit it for every agent. "For this agent", "this app", and "current agent" are explicit scope: read the current name from object_get(ref="") and put that one name in metadata.agents.

  1. Keep the body to procedure, judgment, and traps the agent would otherwise miss. Put repeated

deterministic logic in scripts/, heavy conditional material in references/, and reusable output material in assets/; say exactly when to read each file.

  1. Apply the finished directory as one skill object:

``yaml kind: skill name: concise-brief spec: files: SKILL.md: {from: concise-brief/SKILL.md} references/criteria.md: {from: concise-brief/references/criteria.md} ``

{from: <path>} stores the file content, so the skill outlives the sandbox. Inline short text directly. Fix validation errors and re-apply.

  1. Confirm that the skill was saved and name its audience. A skill with metadata.agents reaches

only those agents; say plainly that another agent will not list or load it. Without that field, every agent whose use_workspace_skills setting holds receives it.

Revise

Use object_get to recover the file digests and the skill's generation, and load_skill to mount the current content. Re-apply the same name with {from: <path>} for changed files, {sha256: <digest>} for unchanged files, and the generation the get returned as a top-level manifest key beside kind, name, and spec. Applying replaces the workspace's copy. Remove it with object_delete using kind: skill.

Traps

  • No blank line or title may precede the opening ---.
  • Quote YAML descriptions; : and similar characters otherwise change their meaning.
  • Bundle only UTF-8 text.
  • Do not reuse a built-in skill name.
  • Never widen "for this agent" to every workspace agent.
  • A save refused for a stale generation means another writer saved first: object_get the skill

again and re-apply from the current state — never retry the same manifest.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
94
Repository forks
12
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/ufo-ai/ufo-core/create-skill/badge)](https://verified-skill.com/skills/ufo-ai/ufo-core/create-skill)

Something wrong with this skill or its page? Tell us.