发现 Skills

Shared project knowledge document that humans and agents can both read and update. Aggregates state from /project-state, /checkpoint, and /memory into a single PROJECT_KNOWLEDGE.md file. Coordinator skill — composes rather than reimplements.

Stress tests the full SPARTA query pipeline: intent classification, QRA retrieval, disambiguation, NLG synthesis, and error detection.

Iterate, evaluate, generate dynamic pages, and push context-aware layouts to Stream Deck.

Unified test runner for skills and packages. Auto-detects sanity.sh, pytest, vitest, lint.sh, and npm test — runs the right framework, returns structured JSON with figure_data for dashboard integration.

Adversarial blind evaluation harness. Generates and runs hidden tests against code written by coding agents. The coding agent never sees the tests — only pass/fail results. Use when /plan creates tasks, or manually with "test-lab run <target>".

N-way concurrent tournament convergence skill. Dispatches N competing strategies via /subagent-service in parallel, scores results against a pluggable metric, diagnoses failures with persona reviewers, /dogpile-researches on combined failure, and iterates until convergence or exhaustion. Manifest-driven (YAML). Two strategies enter, one strategy leaves. Use when: self-improvement loop, tournament, convergence, compete strategies, concurrent training, parallel approaches, best-of-N, arena.

Build TTS datasets and train Qwen3-TTS voice models from audiobooks or curated clips. Use for voice cloning, dataset prep, WhisperX alignment, Qwen3-TTS training, and inference.

🔩 Firmware development for ARM Cortex-M, ESP32, STM32, and AVR — RTOS tasks, communication protocols (I2C/SPI/UART/BLE), power optimization, and OTA updates. Activate for any embedded, IoT, or microcontroller question.

Use when the user asks for actionable engineering advice with clarifying questions, options, tradeoffs, assumptions, and one recommendation.

Run CI-friendly API test suites (REST + GraphQL) from a single manifest, using the bundled api-test CLI and emitting JSON (+ optional JUnit) results. Use when the user asks to reduce CI boilerplate and provide a simple, composable suite runner for other tools (pytest/node/LLM) to call.

Scan a repo using its checked-in Semgrep configuration (local rules only), triage findings, and either fix the most severe issues or open a report-style PR. Use when a repo already has Semgrep config and rules committed; stop if no Semgrep config is present.

A discipline for writing a short, reviewed plan before any non-trivial implementation. Plans live under a configurable plans directory (default `docs/plans/`) and follow a prescribed shape — problem, approach, tradeoffs, steps, open questions. Use when the user says "write a plan", "plan this", or "let's plan".

Research-backed implementation planning skill. This skill should be used when the /plan command is executed, when creating implementation plans, roadmaps, or architectural designs, or when tasks have high uncertainty requiring research before coding. Transforms user requests into execution-ready plans (PRP) using the R.P.I.V workflow (Research → Plan → Implement → Validate).

Route and apply backend and server-side rules across enterprise frameworks. Use for quarkus, vertx, hibernate, guicedee, mapstruct, lombok, spring, jspecify, security-reactive, and fluent-api rule selection.

Route and apply language and framework rules. Use for java, kotlin, angular, react, vue, and typescript topic guidance and compatibility decisions.

Capture and automate macOS UI with the Peekaboo CLI.

Analyze images and generate comprehensive JSON profiles for style recreation. Use when users upload images for visual analysis, style extraction, AI image generation prompts, or need detailed breakdowns of composition, lighting, color, and subject elements.

Test skill for unit tests.

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to 'open a website', 'fill out a form', 'click a button', 'take a screenshot', 'scrape data from a page', 'test this web app', 'login to a site', 'automate browser actions', or any task requiring programmatic web interaction.

Discovery-generated BigQuery v2 API commands — datasets, tables, routines, and models get/list operations.

Expert in the Spec-Kit Plus lifecycle (Specify -> Plan -> Tasks -> Implement). Use this when the user needs to define new features, plan architecture, or decompose work into atomic units.

프로젝트의 마크다운 문서 목록을 TOON 포맷으로 출력합니다

Automate Apex27 tasks via Rube MCP (Composio). Always search tools first for current schemas.

Build robust backtesting systems for trading strategies with proper handling of look-ahead bias, survivorship bias, and transaction costs. Use when developing trading algorithms, validating strategies, or building backtesting infrastructure.