skills

Teach the user one piece of knowledge at a time about a chosen topic, advancing only when they ask, and auto-saving each lesson as a committed markdown note when the working folder is a git repo. Use when the user wants to learn or study a topic, asks to be taught something, is learning a foreign language, or mentions "teach me".

Analyzes a Korean stock by name using the pykrx MCP server and DART, calling tools in order to assemble valuation, financial statements, investor flow, and short-selling data into a single investment-value report. Use when the user asks whether to buy a Korean stock, asks for its investment value, or requests a financial analysis of a listed Korean company (e.g. 삼성전자, 카카오, NAVER).

Scaffolds necessary Next.js App Router file conventions and route-level SEO metadata (page.tsx, layout.tsx, error.tsx, not-found.tsx, loading.tsx, default.tsx, global-error.tsx, metadata, generateMetadata, canonical, robots, openGraph, twitter) for a given route path. Checks existing files, reads them for context, and generates only missing files. Use when user provides a route path and wants to add missing special files or SEO metadata, or asks to "scaffold", "create files for", or "set up route files" for a Next.js App Router route.

Builds AI music generation prompts through an interview-driven songwriting workflow, producing Custom Mode-ready Title, Style, and Lyrics blocks. Use when the user asks to make, write, compose, prompt, structure, or refine a song, track, beat, instrumental, lyrics, or AI music prompt.

Review knowledge already saved in the current repo by quizzing one item at a time with spaced-repetition scheduling, keeping all session state in a throwaway dot-folder that is deleted when done. Use when the user wants to review, quiz, or test themselves on previously learned notes, or says "review me".

Run a short adaptive diagnostic to gauge how much the user already knows about a topic before learning it, then save a prose level summary to a single git-tracked levels.md. Use when the user is about to learn a topic and wants a level check first, asks to be tested on what they know, or says "place me".

Adds an optional HTML-first visual reference for brainstorming with mockups, diagrams, wireframes, and side-by-side visual choices. Use when a discussion would be easier to resolve visually, especially for UI layouts, flows, architecture diagrams, spatial relationships, visual hierarchy, or design direction comparisons.

Create a new specification file for the solution, optimized for Generative AI consumption.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree, and drawing HTML visual references when a question is easier answered visually. Use when user wants to stress-test a plan, get grilled on their design, or mentions "step by step".

Stress-test and refine user prompts until they are specific, unambiguous, and execution-ready. Use when the user asks to improve, validate, critique, rewrite, or make a perfect prompt, or when they mention prompt quality, ambiguity, missing context, unclear requirements, or prompt engineering.

Build a brand's AEO/GEO (Answer Engine Optimization / Generative Engine Optimization) strategy step by step over a 3-phase roadmap. Generates measurement spreadsheets, a share calculator that emits dated markdown reports, runs an interview to build a canonical entity definition, scaffolds JSON-LD Organization schema with stack-specific injection guides, and tracks monthly trend in a single accumulating CSV. Use when user wants to set up AEO/GEO, optimize brand mentions in ChatGPT/Claude/Gemini answers, measure AI answer share, build entity definition or schema markup for AI visibility, or asks "how do I get AI to recommend my brand", "AEO 셋업", "GEO 전략 만들어줘", "AI 답변 점유율 측정".

Build agent harness infrastructure (Constrain, Inform, Verify, Correct) on a project step by step. Generates AGENTS.md as the canonical rules file with a CLAUDE.md import stub, configures Claude Code permissions and a session-end verify hook, scaffolds optional CI gates, and seeds a Past Failures section for the "1 failure = 1 prevention rule" loop. Optimized for Claude Code; AGENTS.md is the cross-tool standard so OpenAI Codex reads the same file. Use when user wants to set up agent guardrails, harden an AI coding agent setup, write or improve AGENTS.md/CLAUDE.md, configure agent permissions/hooks, or asks "how do I make my agent safer", "set up harness engineering", "configure agent rules for this project".

Search npm packages by functionality using the npm registry API, iteratively refining the query until the user finds the right one. Covers JavaScript and TypeScript packages only — does not cover pip, cargo, gems, or other package registries. Use when user wants to find an npm package, node module, or JavaScript/TypeScript library for a specific purpose, or says things like "find me an npm package for X", "is there a node module that does Y", or "recommend a JS library for Z".

Search GitHub repositories by functionality using gh CLI, iteratively refining the query until the user finds the right one. Use when user wants to find a GitHub repository, library, or tool for a specific purpose, or says things like "is there a repo that does X", "find me a library for Y", or "recommend a tool for Z".