skill-trigger-metadatalisted

# Skill Trigger Metadata Bundled commands use `$PLUGIN_ROOT` (`$env:PLUGIN_ROOT` in PowerShell; same path suffix) for the plugin root. Set it once: use the host's plugin-root variable when defined (Claude Code: `PLUGIN_ROOT="$CLAUDE_PLUGIN_ROOT"`), otherwise the absolute path of this skill folder's `../..`. Design `name` and `description` as the retrieval contract that makes an agent read the right `SKILL.md`. Use this before finalizing skill frontmatter or plugin skill metadata. If the task also changes skill instructions, resources, scripts, packaging, or installation, use this skill for metadata first, then use `skill-factory` or `plugin-factory` for the broader lifecycle. ## Core Rule The description must answer: should the agent read this skill now? Write for progressive disclosure: `name` is the searchable handle, `description` is the trigger contract, and `SKILL.md` body contains workflow, commands, constraints, and examples. Do not put enough workflow detail in `description` for the agent to act without reading `SKILL.md`. ## Trigger Selection Model Treat skill selection as tool lookup plus routing: - information scent: metadata gives strong proximal cues that the full skill is worth reading; - local vocabulary: include task terms, artifacts, file types, symptoms, and synonyms from the real work surface; - tool decision vector: say when to use the skill, expected inputs, adjacent non-use cases, and failure symptoms; - negative boundaries: include near-miss p