skill-doctorlisted

# skill-doctor — Agent Skill Analysis via CLI `skill-doctor` is a **local-only** CLI (it never uploads your skills) that scans the user's installed AI-agent skills across platforms (Claude Code, Cursor, Copilot, Codex, Gemini CLI, Windsurf, …) and reports duplicates, conflicts, security risks, and context-cost. This skill wraps it: **you (the agent) drive the CLI, read its `--json` output, and produce the analysis.** ## Core principle (read first) - **The CLI is the source of truth.** Run `skill-doctor` commands and use their `--json` output as your reasoning input. Do not re-implement discovery yourself. - **Do NOT read raw SKILL.md files** of the user's installed skills to perform analysis. `skill-doctor` already discovers, parses, and structures them. Only read a specific raw skill file when the user **explicitly** asks ("show me the raw SKILL.md of X") or requests a targeted edit. - Keep it local: never upload or exfiltrate the user's skills. ## Step 0 — Mode detection (run every invocation) 1. **Is the CLI present?** `skill-doctor --version`. If it fails, tell the user to install (`npm i -g @evilstar2025/skill-doctor`, Node ≥ 20) and stop. 2. **Is a backend model configured?** Run: ``` skill-doctor config view --json ``` Inspect `analysis.apiKeyConfigured` and `embedding.apiKeyConfigured`. - Both configured → **ENHANCED mode** (CLI can do LLM grouping, LLM audit, semantic conflict detection). - Neither configured → **FALLBACK mode**