# Skill Advisor > Recommend which of the user's ALREADY-INSTALLED skills best fits a task or query. Use when the user asks "what skill should I use for X", "which of my skills fits this", "do I have a skill for Y", "what's at my disposal for Z", "recommend a skill from my toolkit", or when you need to pick the right skill from their available toolkit before starting work. This skill is read-only — it advises, it does not install. For installing new skills, use the separate skill-management workflow available in the current agent environment. - Skill: `kennykankush/skill-advisor` (Agent Skill) - Install (CLI): `npx skillmds@latest add kennykankush/skill-advisor` - Raw SKILL.md: https://api.skillmd.com/api/skills/kennykankush/skill-advisor/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: kennykankush (https://skillmd.com/u/kennykankush) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/kennykankush/skill-advisor --- # Skill Advisor — Recommend from the User's Installed Toolkit Your job is to help the user pick the right skill from what's **already installed**. You don't install anything. You don't do web research. You read their local skill index and match it to the task at hand. This skill activates on queries like: - "What skill do I have for X?" - "Which of my skills should I use here?" - "Recommend from my toolkit" - "Do I have something for Y?" - "What's at my disposal for Z?" ## How to answer ### Step 1 — Understand the task Identify the domain and specifics. Examples: - "build a landing page" → frontend design + motion + polish - "make an iOS animation" → iOS + animation - "review this component" → critique/audit flavor - "set up a Neon database" → backend/data ### Step 2 — Read the index Always start by reading `~/.agents/CATEGORIES.md` — it's the canonical categorized view of every installed skill, regenerated automatically after changes. If `CATEGORIES.md` doesn't exist or is stale, fall back to reading the relevant local registries for the current agent environment: - `~/.codex/plugins/cache/` (Codex plugin cache) - `~/.agents/skills/` (shared local skills) - `~/.agents/.skill-lock.json` (flat skills via npx skills) - `~/.claude/plugins/installed_plugins.json` + plugin caches (plugin skills) - `~/.claude/skills/` (standalone / skillfish-managed) ### Step 3 — Match skills to the task Go through each relevant category in `CATEGORIES.md`. For each candidate skill, ask: - Does its description trigger on THIS task? - Is it a good fit or just tangentially related? - How does it stack against similar skills in the same category? ### Step 4 — Rank and recommend Give a ranked list (2-5 skills max — more is noise). For each: - Full namespaced name (e.g., `impeccable:shape`, not just `shape`) - One-line reason it fits - When in the workflow to use it (pre-work, build, polish, review) ### Step 5 — Flag gaps If the user's request reveals a category where their toolkit is THIN (e.g., they ask for Android UX and only have iOS skills), say so: > "For Android specifically, you currently have `ui-design:mobile-android-design` but not much else. Consider asking `skill-manager` to discover more Android-focused skills." This is how `skill-advisor` hands off to `skill-manager` naturally. ## Response format Use this structure: ``` 📚 For , from your installed toolkit: 🥇 **** — 🥈 **** — 🥉 **** — Optional: also consider **** for . Workflow suggestion: 1. Start with X to plan/scaffold 2. Use Y during build 3. Apply Z before shipping Gap: ``` Keep it tight. No preamble. No "I'll help you with that" fluff. ## What NOT to do - **Don't install anything.** If the user needs a new skill, say "ask skill-manager to discover/install X." - **Don't do web research.** Local data only. - **Don't recommend project-scoped skills when the user is in a different project.** Respect scope. - **Don't recommend skills the user doesn't have.** Only recommend from CATEGORIES.md. - **Don't be verbose.** 3 ranked recommendations beat 10 loose mentions. ## Toolkit grounding The toolkit changes constantly — skills get installed, removed, and migrated across mechanisms (plugins, npx skills, skillfish, hand-authored). `CATEGORIES.md` and the live registries are the **only** source of truth. Never recommend from a remembered or assumed inventory; what was installed last month may be gone, and what's installed now may not be in any prior snapshot. Project-scoped skills exist in some repos. Only recommend those inside their own project — respect scope. ## Example exchanges **User:** "Which of my skills should I use for a landing page?" **Response:** ``` 📚 For a landing page, from your installed toolkit: 🥇 **impeccable:shape** — plan the UX/UI before coding (discovery interview → design brief) When: before you write any code 🥈 **ui-design:visual-design-foundations** — type, color, spacing fundamentals When: during initial design decisions 🥉 **frontend-design:frontend-design** — distinctive production-grade interface implementation When: during build Also consider: **impeccable:bolder** if the page feels safe/boring, and **impeccable:polish** before shipping. Workflow suggestion: 1. impeccable:shape to plan 2. ui-design:visual-design-foundations + frontend-design:frontend-design during build 3. impeccable:polish before ship Gap: none for this task — you're well-covered. ``` **User:** "what do I have for Android?" **Response:** ``` 📚 For Android UX specifically: 🥇 **ui-design:mobile-android-design** — Material Design patterns for Android That's it for Android-specific skills in your toolkit. You have strong iOS coverage (ios-design, ui-design:mobile-ios-design) and cross-platform (ui-design:react-native-design), but Android alone is thin. Gap: If you want more Android depth, ask skill-manager to discover Android-focused skills. ```