skill-authoring

# Skill Authoring ## Purpose Write a skill that loads at the right time and materially changes the agent's behavior when it does. A skill that never triggers is dead weight; a skill that triggers constantly is a system prompt with extra steps. ## When to Use - Creating a new skill. - A skill that does not activate when it should, or activates when it should not. - Converting a recurring prompt into a reusable capability. - Reviewing a skill before publishing it. ## Capabilities - Scoping: what belongs in a skill and what does not. - Description writing, which is what determines triggering. - Structuring the body for an agent to act on. - Progressive disclosure through supporting files. - Testing that the skill triggers on real requests. ## Inputs - The capability the skill should provide. - The requests that should trigger it, phrased as a user would phrase them. - The requests that should *not* trigger it. ## Outputs - A `SKILL.md` with valid frontmatter and an actionable body. - A description that triggers reliably on the intended requests. - Supporting files for detail that is not always needed. ## Workflow 1. **Scope it to one capability** — A skill that covers "backend development" is too broad to trigger precisely and too shallow to be useful. A skill about API design is neither. 2. **Write the description last, and carefully** — It is the only part the agent sees before deciding whether to load the skill. It must state what the skill does *and* when to use ...