core-auditlisted

# Audit Analyze a skill from the current conversation for quality. Two modes: quick (analysis only) or extended (internal walkthrough + analysis). Refactor proposals are presented as a plan via plan mode for single-shot approval. **Trigger**: `/core-audit` ## Step 1: Load Skill from Chat Scan the conversation above for unique skill invocations (slash commands like `/dev-build`, `/core-edit`, etc., or skill names referenced in `<command-name>` tags). **Resolution rules:** (Note: `core-audit` itself is always present in the conversation via the `<command-name>` tag, so at least one skill is always detected.) - **Exactly one unique skill** → auto-select. Show: ``` AUTO-SELECTED: [name] (only skill in conversation) ``` Proceed to load. - **Two or more unique skills** → use **AskUserQuestion**: - header: "Skill" - question: "Which skill from this conversation do you want to audit?" - options: one per detected skill, in order of most-recent invocation first; the first option gets "(Recommended)" appended to its label - multiSelect: false Do NOT add a "list all skills" or "other" option — only the skills used in this chat are eligible. **After resolution, load** `.claude/skills/[name]/SKILL.md` and show: ``` LOADED: [name] [one-line summary from description] Sections: [count] | Has resources: [yes/no] ``` ## Step 1.5: Context Detection Scan for prior skill execution that can inform testing: 1. **Conversation context** — the invocation that triggered