skills

Use for implementation, review, refactoring, migration, or diagnosis in a React 18 or React 19 project built with Vite 7 when the task requires React design judgment, version compatibility, or a Vite client build decision. The primary scope is browser-rendered CSR SPA/MPA and Vite client builds embedded in an existing backend. Do not use by default for non-Vite React, React 17 or earlier, React 20 or later, Vite 6 or earlier, Vite 8 or later, framework-owned SSR/RSC/server functions, or isolated copy and styling work.

Generate an executable implementation work-plan Markdown at `docs/briefs/` from planning notes or a rough task description, with the plan body written in caveman **full** mode as a register-only transform that preserves every fact, bullet, execution stage, field, and section depth from the normal-mode equivalent. Work brief, task brief, handoff brief, implementation ticket, and task spec remain trigger aliases. Nine required sections keyed to Conventional Commits types so coding agents switch behavior (refactor → preserve, fix → reproduce first, perf → measure first). Briefset mode emits a parent execution-management document plus N child briefs when the input describes multiple execution contexts. Chat / questions / status reporting stay in normal prose — caveman applies to the saved file only. Use only when the user explicitly asks for the caveman register or invokes this skill by name; for normal-prose briefs use task-brief-creator instead.

Generate an executable implementation work-plan Markdown at `docs/briefs/` from planning notes or a rough task description. Work brief, task brief, handoff brief, implementation ticket, and task spec remain trigger aliases. Nine required sections are keyed to Conventional Commits types so coding agents switch behavior (refactor → preserve, fix → reproduce first, perf → measure first). Briefset mode emits a parent execution-management document plus N child briefs when the input describes multiple execution contexts. Manual trigger only — use when the user explicitly invokes this skill or asks for a work plan, implementation plan, work brief, task brief, handoff brief, implementation ticket, or task spec for a coding agent. Not for prose summaries, status reports, design docs, or meeting notes. For the plain-language caveman variant, use task-brief-creator-caveman instead.

Use only when the user explicitly invokes this skill.

Analyze repository structure and generate or update standardized AGENTS.md files that serve as contributor guides for AI agents. Supports both single-repo and monorepo structures. Measures LOC to determine character limits and produces structured documents covering overview, optional ownership maps, patterns, conventions, and working agreements. Update mode refreshes only the standard sections while preserving user-defined custom sections, and only modifies AGENTS.md files previously generated by this skill — externally-authored files are reported, never rewritten. Use when the user wants an AGENTS.md generated or refreshed — setting one up for a new repository, onboarding AI agents to an existing codebase, updating an existing AGENTS.md after meaningful project changes, or when the user mentions AGENTS.md. Not for CLAUDE.md, README, or other documentation files.

Orchestrate a large task as a manager-run sequence of isolated sub-agent phases when the user explicitly invokes this skill or asks for this. The main agent decomposes the task into phases of different kinds, picks the topology that fits (linear pipeline, parallel fan-out, dependency DAG, or iterative loop), assigns each phase a capability-by-deliberation tier, and runs each as a fresh isolated sub-agent exchanging file paths only -- never doing phase work itself. It proves any expensive step cheaply before committing, and halts for the user at up to two points: go/no-go before that commitment, and accept/reject on review findings. Runs are resumable by run id; a changed condition re-versions the prior run. For large heterogeneous work: research sweeps, multi-variant runs, migrations, audits. Explicit-invocation only. Not for single-artifact build-review-improve loops (use delegated-review-loop) or single-answer blind verification (use iterative-self-review).

Create, update, or fact-check a Feature Design Doc that captures product/system design decisions for downstream code agents. Use whenever the user mentions "기능 문서", "기능 설계 문서", "Feature Design Doc", "FDD", "design doc", or asks to write, draft, generate, update, revise, review, validate, fact-check, audit, or normalize a feature document — even when the user does not explicitly use the word "design", and especially right after a feature was ideated and implemented in-session (vibe-coding) and now needs documenting. Trigger on near-misses like "기능 정리 문서 좀 써줘", "이 문서 좀 검토해줘 우리 기능 design 관련이야", or "design doc 좀 만들어볼래". NOT for system/architecture design docs, RFCs, implementation plans, task lists, or READMEs — the subject must be a single product feature. This skill enforces that the artifact remains a decision source, not an implementation plan or task list, so use it whenever the working artifact is meant to drive what a feature *is* rather than *how it gets built*.

A skill that analyzes user requirements to generate system prompts ready for evaluation. It determines whether a single or multi-prompt architecture is needed and queries for missing information if requirements are insufficient. Use when asked to create, design, or draft a system prompt for an LLM-powered bot, assistant, classifier, router, or multi-step prompt pipeline (e.g. "create a system prompt", "시스템 프롬프트 만들어줘", "분류기 프롬프트 짜줘"). Not for refining an existing prompt, one-off user prompts, agent definition files, CLAUDE.md/AGENTS.md, or Claude Code skill definitions.

Run a delegated build-review-improve loop when the user explicitly invokes this skill or explicitly asks to have sub-agents build and iteratively review the work. A worker sub-agent produces the artifact (with an optional scout sub-agent gathering context first), a task-matched panel of independent reviewer sub-agents critiques it in parallel, and the worker improves until the panel passes or the iteration cap is reached. All inter-agent content flows through documents in a per-session directory; the main agent orchestrates by passing file paths only and never writes the artifact itself. Applies to any deliverable: code changes, documents, designs, plans. Do not auto-trigger on ordinary requests; this is an explicit-invocation skill. Distinct from iterative-self-review, where the main agent holds the pen and a single blind verifier checks it.

IntelliJ Platform plugin development for JetBrains IDEs. Use when writing, debugging, or migrating a JetBrains plugin — `plugin.xml`, services, actions, PSI/VFS/Document, EDT/BGT threading and Read/Write actions, Kotlin coroutines on the IDE platform, custom languages (Grammar-Kit, JFlex, lexer/parser, syntax highlighting, completion, references), code insight (folding, inspections, intentions, inline completion, inlay hints), Kotlin UI DSL v2, tool windows, settings, run/debug configurations, External System (Gradle/Maven), IntelliJ Platform Gradle Plugin 2.x, dynamic reload, Plugin Verifier, signing, and Marketplace publishing; legacy migrations (`ApplicationComponent`, `getCoroutineScope`, raw `Thread`/`ExecutorService`). Not for end-user IDE configuration or vmoptions tuning.

release-it configuration, setup, and plugin development. Triggers on .release-it.* config files, release-it CLI usage, version bumping, changelog generation, npm publishing, GitHub/GitLab releases, git tagging, hooks lifecycle, pre-release workflows, CI/CD pipelines, monorepo strategies, and custom plugin development.

Create structured Linear issues (main + sub-issues) with project linking, title prefix, and labeling. Supports two workflows: Generic (code tasks) and PRD Pipeline (제품 요구사항 정의서). PRD Pipeline applies content-strategy, content-production, content-humanizer, copy-editing principles to produce high-quality issue descriptions targeting fullstack developers in Korean. Use when asked to create, register, or break down tasks into Linear issues. Not for querying or updating.

Review completed Linear sub-issues: cross-validate Done Criteria, code changes, and worker comments to approve, request changes, or ask for clarification. Use when asked to review, validate, or QA done Linear issues.

Execute code tasks from Linear sub-issues: resolve dependencies, implement changes, validate Done Criteria, and sync status. Use when asked to work on, implement, or start coding a Linear issue (e.g. PRI-42).

Provides JetBrains IDE VM options knowledge for version-specific GC selection and memory/performance tuning (JDK 17/21, IDE 222+). Use this skill whenever the user mentions JetBrains IDE performance, vmoptions, .vmoptions files, IntelliJ/WebStorm/PyCharm/GoLand/Rider/CLion/PhpStorm/RubyMine/DataGrip tuning, IDE freezes or lag, heap size configuration, GC tuning for IDEs, or wants to optimize their JetBrains IDE startup or runtime performance — even if they don't explicitly mention "vmoptions".

Guide users through a structured 3-stage workflow for co-authoring documentation — Context Gathering, Refinement & Structure, and Reader Testing. Use this skill whenever the user wants to write documentation, proposals, technical specs, decision docs, RFCs, PRDs, design docs, or any structured content. Trigger when the user mentions writing docs, drafting a proposal, creating a spec, writing up decisions, or similar documentation tasks, even if they just say "write a doc" or "let's draft something."

Performs production-ready code reviews on git changes. Supports commit/range/file-scoped analysis, impact assessment, breaking-change detection, confidence-aware finding classification, and risk-weighted verdict generation.

Generates structured Google Veo 3.1 video prompts by collecting user input for subject, action, style, cinematography, and audio. Guides users through the Universal Prompt Formula to produce camera-ready prompts. Use when creating Veo video prompts or when the user mentions Veo, video generation, or cinematic prompts.

Performs OWASP-based code security audits on any codebase. Analyzes source code against ASVS 5.0.0 verification requirements, API Security Top 10 2023 risk patterns, OWASP CheatSheet secure coding practices, and WSTG testing methodologies. Input is a codebase to review; output is a detailed Markdown security audit report. Use when the user requests a security audit, security review, vulnerability assessment, or code security analysis.

Generate Typst source code for documents, reports, papers, and presentations, covering markup, math, scripting, and layout syntax.

Identify and validate profitable business opportunities by analyzing market size (TAM/SAM/SOM), unit economics, competitive landscape, and PMF indicators. Generates comprehensive HTML reports with opportunity scorecards.

Capable of converting raw SQL queries into type-safe Kysely TypeScript code. Knows how to handle various SQL dialects and complex query structures.