| 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 | [object Object] |
Yao Meta Skill
Build reusable skill packages, not long prompts.
Router Rules
- Route by frontmatter
descriptionfirst. - Keep
SKILL.mdto routing plus a minimal execution skeleton. - Put long guidance in
references/, deterministic logic inscripts/, and evidence inreports/. - 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
- Decide whether the request should become a skill, then choose the lightest archetype.
- Begin with a short, human intent dialogue to capture the recurring job, outputs, trigger phrases, exclusions, constraints, and the user's taste or standards.
- 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.
- Write the
descriptionearly, then test route quality before expanding the package. - 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. - 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
2-3high-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:
- acknowledge the seed idea
- explain that the goal is to shape a reusable skill around the real work and desired outcome
- invite a natural reply first
- 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 / Exclusionsas 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:
- a working skill directory
- a trigger-aware
SKILL.md - aligned
agents/interface.yaml - optional
references/,scripts/,evals/,reports/, andmanifest.jsononly when justified - 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.