Yao Meta Skill

Build reusable skill packages, not long prompts.

Router Rules

• Route by frontmatter description first.
• Keep SKILL.md to routing plus a minimal execution skeleton.
• Put long guidance in references/, deterministic logic in scripts/, and evidence in reports/.
• Use the lightest process that still makes the skill reliable.

Modes

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

Mode rules: Operating Modes, QA Ladder, Resource Boundary Spec, Skill Engineering Method.

Compact Workflow

1. Decide whether the request should become a skill, then choose the lightest archetype.
2. Begin with a short, human intent dialogue to capture the recurring job, outputs, trigger phrases, exclusions, constraints, and the user's taste or standards.
3. Run a short reference scan with high-quality external benchmark objects first, then ask whether the user has references worth learning from; use local files only for fit, privacy, and compatibility calibration.
4. Write the description early, then test route quality before expanding the package.
5. Add only the folders and gates that earn their keep: trigger_eval.py, optimize_description.py, judge_blind_eval.py, resource_boundary_check.py, governance_check.py, cross_packager.py.
6. After the first package exists, surface the top three next iteration directions instead of expanding the skill in every direction at once.

Core playbooks: Method, Intent Dialogue, Reference Scan, Archetypes, Gate Selection, Iteration Philosophy, Non-Skill Decision Tree, Eval Playbook.

First-Turn Style

When the skill first activates, do not open with a bureaucratic intake form.

• Mirror the user's language.
• Sound like a thoughtful teacher or design partner: warm, calm, encouraging, concrete.
• Start by helping the user feel understood before asking for structure.
• Ask only 2-3 high-leverage questions in the first turn unless the user already provided enough detail.
• Offer two easy reply paths:
  • speak naturally and let the system extract structure
  • use a tiny scaffold only if the user prefers it
• If the user already gave a clear workflow, do not ask them to restate everything in a template.
• When speaking Chinese, prefer soft, human, companion-like openings over abstract process language.

Preferred opening shape:

1. acknowledge the seed idea
2. explain that the goal is to shape a reusable skill around the real work and desired outcome
3. invite a natural reply first
4. only then offer a lightweight template as an optional shortcut

Avoid this failure pattern:

• dumping a cold field list such as Name / One-line ability / Inputs / Outputs / Exclusions as the default first reply
• sounding like a form collector instead of a guide
• asking for architecture before understanding the human job to be done

Output Contract

Unless the user asks otherwise, produce:

1. a working skill directory
2. a trigger-aware 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, benchmark objects, gates, and next steps

Reference Map

Primary references: Method, Reference Scan, Intent Dialogue, Archetypes, Gate Selection, Iteration Philosophy, Governance, Resource Boundaries, Eval Playbook.