What problem does it solve? Writing a domain-expert persona for the ADD engine is easy to get wrong: a persona that misses the engine's contract is never loaded, and one that duplicates the agent's own voice or output contract is dead weight. This Skill guides authoring, seeding, and validating persona files so they route correctly and carry real judgment. ## Core Features & Use Cases - Schema-guided authoring: Produces personas with the required sections (Identity, Critical Rules, Default Requirement, Success Metrics) plus recommended frontmatter (flow, task-kinds, use-when, not-when) that the ADD engine validates and routes on. - Seeding from existing sources: Distills a first draft from the teacher persona library or a sample subagent file instead of starting from a blank page, with explicit source-to-schema mappings. - Judgment-layer patterns: Applies twelve distilled patterns (earned-perspective identity, bold-lead rules, qualification gate, failure-aware metrics, escalation stances) that separate an expert lens from a keyword list. - Use Case: When no existing persona fits a new task kind (e.g., a payments-integration lens), use this Skill to seed from the nearest teacher persona, write the judgment sections, and validate with add doctor until the persona appears in the compiled roster. ## Quick Start Use the persona-author skill to draft a new ADD persona for payment integration work, seeded from the nearest teacher persona, and validate it with add doctor.
