resonance-ops-skill-author

Write the description in third person, state what the skill does and when to use it, and lead with concrete trigger terms a user would say. Keep it under 1024 characters; the description is the only text loaded at startup for skill selection.

Choose the procedure archetype, write at least 3 eval cases first, then build from the matching template preserving Input, Output, Definition of Done, and Recovery sections. Compile with forge.py and run validate_skill.py before committing.

Knowledge skills are domain experts applied inline and auto-loaded. Procedure skills are gated multi-step jobs invoked as /name with a Definition of Done. Orchestration skills are procedures that drive other skills or subagents in an ordered pipeline.

No. Generated SKILL.md files carry a do-not-edit banner. Edit the template under .forge/skills/<name>/ and recompile with forge.py so injected sections like voice and completion stay consistent.

Common causes are a description over 1024 characters, missing Input/Output/Definition of Done/Recovery sections in procedure skills, or an invalid name. The skill is not done until validate_skill.py reports clean and evals pass.

Skip it when you cannot name a concrete gap with an example task, when evals show no improvement over the no-skill baseline, or when the request is vague like making output more elegant. An unmeasured skill adds tokens without value.