skill-authoringlisted

## Routing Guidance - WORKFLOW mode activates when user says "build a skill", "make a skill for X", "write a skill", "create a new skill", "add a skill", or "I need to author a skill". - METHODOLOGY mode activates when user asks to "optimize skill descriptions", "benchmark triggers", "improve routing", "fix trigger issues", "test if my skill fires", or "fix why my skill doesn't load". - Do NOT use for general code writing or project planning. - Use `context: fork` when this skill orchestrates multiple steps, produces heavy parallel output, or builtin agent behavior matches the goal. - Do NOT use when the skill is simple (single tool call), you need intermediate results in main conversation, or builtin behavior contradicts this skill's purpose. ## Decision Router IF creating a new skill within this plugin's structure → use **WORKFLOW** mode IF writing descriptions that trigger reliably → use **METHODOLOGY** mode IF optimizing existing trigger performance → use **METHODOLOGY** mode --- ## WORKFLOW MODE *Creating new skills in this plugin's structure* ### Pre-Flight Validation Checklist Before committing a skill, verify: - [ ] Description triggers on intended inputs (test with `claude -p`) - [ ] Description does NOT trigger on off-topic inputs (expect 0) - [ ] Description ≤ 150 chars - [ ] when_to_use ≤ 200 chars - [ ] Body ≤ 500 lines - [ ] No invalid frontmatter fields (metadata, related_skills, tags) - [ ] Skill name is kebab-case, max 64 chars, no XML tags, no "an