---
name: yao-meta-skill
description: Create, refactor, evaluate, and package agent skills from workflows, prompts, transcripts, docs, or notes. Use when asked to create a skill, turn a repeated process into a reusable skill, improve an existing skill, add evals, or package a skill for team reuse.
metadata:
  author: Yao Team
  philosophy: "structured design, evaluation loop, template ergonomics, operational packaging"
---

# Yao Meta Skill

Build reusable skill packages, not long prompts.

## Router Rules

- Route by frontmatter `description`.
- Keep `SKILL.md` lean.
- Put guidance in `references/`, logic in `scripts/`, and evidence in `reports/`.
- Use the lightest reliable process.

## Modes

- `Scaffold`: exploratory or personal use.
- `Production`: team reuse with focused gates.
- `Library`: shared infrastructure or meta skill.

Mode rules: [Method](references/skill-engineering-method.md), [Operating Modes](references/operating-modes.md), [Resource Boundaries](references/resource-boundaries.md).

## Compact Workflow

1. Decide whether the request should become a skill and choose the lightest fit.
2. Capture job, output, exclusions, constraints, and standards.
3. Run a reference scan: external benchmarks first, user references second, local fit third. Keep synthesis silent unless intent stays unclear or a real design conflict needs a user call.
4. Write the `description` early and test route quality before expanding the package.
5. Add output-risk and artifact-design profiles when user-facing deliverables matter.
6. Add only folders and gates that earn their keep.
7. Surface the top three next iteration directions.

Core playbooks: [Method](references/skill-engineering-method.md), [Intent Dialogue](references/intent-dialogue.md), [Reference Scan](references/reference-scan.md), [Artifact Design](references/artifact-design-doctrine.md).

## First-Turn Style

When the skill first activates:

- open warmly, like a thoughtful teacher or design partner
- start from the user's work and desired outcome before asking for structure
- ask only `2-3` key questions unless the user already gave enough detail
- let the user answer naturally first; offer a tiny scaffold only as a shortcut
- avoid cold field lists and turn benchmark work into one recommendation unless uncertainty or conflict needs a visible call

Chinese conversations should sound soft and companion-like rather than procedural.

Opening patterns: [Intent Dialogue](references/intent-dialogue.md).

## Output Contract

Unless the user asks otherwise, produce:

1. a working skill directory
2. a `SKILL.md`
3. aligned `agents/interface.yaml`
4. optional `references/`, `scripts/`, `evals/`, `reports/`, and `manifest.json` only when justified
5. a short summary of boundary, exclusions, gates, and next steps

## Reference Map

Primary references: [Method](references/skill-engineering-method.md), [Reference Scan](references/reference-scan.md), [Artifact Design](references/artifact-design-doctrine.md), [Output Visual Quality](references/output-visual-quality.md), [Governance](references/governance.md).