Use when a requirement or an unproven technique must be made clickable before it is built — produces a single self-contained HTML prototype that doubles as a design spike, whose control panel is the open questions. Rebuilds existing screens at high fidelity from real source (structure, copy strings, icons, tokens, each cited file:line; invents only what the requirement adds), derives one control per acceptance criterion or technical unknown, drives real data through a dev-server proxy that also proposes the route shape (live > capture > labeled stub, never invented), verifies every state in a real browser, and reports what the spike settled. Triggers on 'prototype FR-NNN', 'clickable prototype of FR-NNN', 'make this requirement clickable', 'spike this technique'. NOT for inventing UI from a prompt — use a generative-design skill. NOT for production frontend code. NOT the client-confirmation record — that is a spec-playback step. NOT an ADR writer.

Remove AI-writing tells and rewrite prose to sound natural and human. Triggers: humanise, humanize, de-AI, deAI, sound natural, less robotic, remove AI tells, AI writing, chatbot tone, Wikipedia AI writing signs, voice match, voice calibration. Uses AskUserQuestion for intake; Read/Write for files; WebSearch/WebFetch when the user wants terminology or usage research; Bash when applying edits in a repo. Outputs: interactive Q&A then rewritten text (and optional brief edit log). Do NOT use for: inventing facts, changing the author's stance, weakening safety or compliance language without explicit instruction, or skipping intake when the user's goals and preservation rules are unclear.

Contribute to ai-slop-gate — an open-source CI/CD policy gate that detects AI-generated code slop, security issues, and compliance violations. Use when adding a new LLM provider (Gemini, Groq, Ollama) or static provider, writing analysis rules, extending the policy engine, working with reporters, fixing compliance sidecar logic, or writing tests. Activate when working with ai_slop_gate/ source tree, policy.yml, snapshot, or GitHub Actions workflows.

Adversarial reviewer personality for architecture discussions. Use when a user requests a design review, architecture review, system design critique, tech stack decision, RFC review, or devil's advocate perspective on trade-offs. Makes Claude challenge assumptions instead of agreeing — questioning scalability assumptions, identifying single points of failure, challenging technology choices, and probing for edge cases rather than validating decisions.

Execute a strict Red-Green-Refactor TDD cycle for a single requirement at a time, in any language or framework. Use when the user provides a business rule, acceptance criterion, or feature requirement and wants a failing test written first, followed by minimum implementation, then a clean refactor. Works for unit tests, integration tests, UI component tests, and API tests across stacks including TypeScript, JavaScript, React, .NET (C#/xUnit/NUnit), Python, Java, and others. Triggered by phrases like "write a test first", "TDD", "red-green-refactor", "behavioral test", "test-driven", "failing test first", or "write the test before the code".

Health check and readiness endpoints for web services — liveness probes, dependency checks, and graceful startup/shutdown. Use when building any backend service that will be deployed behind a load balancer, in containers, or in any environment that monitors service health.

Vue 3 Composition API patterns that agents commonly get wrong: reactivity pitfalls, composable structure, Pinia setup stores, watch vs computed, storeToRefs, template refs, provide/inject, and accessibility. Use when building or reviewing Vue 3 apps.

SQLite best practices for Node.js with better-sqlite3 — WAL mode, pragma configuration, foreign keys, transactions, migrations, STRICT tables, prepared statements, graceful shutdown, and performance patterns. Use when building or reviewing a Node.js app with SQLite, when you see better-sqlite3 imports, when setting up a new database, when debugging slow queries or SQLITE_BUSY errors, or when migrating from in-memory storage to SQLite.

Integration tests for ASP.NET Core APIs using xUnit and WebApplicationFactory. Covers project setup, service replacement with ConfigureTestServices, database isolation, the 5 essential test patterns, authentication testing, and FluentAssertions. Use when a .NET API has no tests, when adding endpoints, or when a reviewer asks for test coverage.

Build chat applications with PubNub Chat SDK

Authors property-based tests in JavaScript / TypeScript using fast-check - wires `fc.assert(fc.property(arbitrary, ...))`, picks arbitraries (`fc.integer`, `fc.string`, `fc.array`, `fc.tuple`, `fc.record`), uses `.map()` / `.chain()` / `.filter()` to build domain arbitraries, and integrates with Jest / Vitest / Mocha / Jasmine / AVA / Tape. Use when a JS/TS codebase needs PBT to catch edge cases - fast-check has been used to find bugs in major libraries (`query-string`, etc.) and is trusted by Jest, Jasmine, fp-ts, Ramda.

Enforces Conventional Commits format and branch naming conventions, validating commit message structure (type/scope/description header), suggesting branch name patterns (feature/*, fix/*, hotfix/*), and enforcing breaking change notation. Use when the user is about to commit, creating a branch, reviewing commit history, preparing a pull request, setting up commit linting, or asking about commit message format, branch naming, or Conventional Commits.

Use when hand-writing, generating, or editing Android Tasker XML files (.tsk.xml, .prf.xml, .prj.xml, .scn.xml, or a full backup .xml) to define tasks, profiles, projects, or scenes for import into the Tasker app — covers the TaskerData node skeleton, numeric action codes, and the per-argument Str/Int/App encoding.

One-thread-per-device async controller for rate-limited IoT APIs with min-interval throttle, stability filter, and send-latest semantics. Use when a real-time producer (camera loop, sensor feed, event stream) drives a cloud or LAN IoT device that can't keep up with per-frame updates, or when you see flicker / HTTP 429 errors from hammering an actuator.

Renders a segmented progress bar that fills bottom-up with a red/yellow/green gradient — the thermometer pattern users expect, not top-down list order. Use when the user asks for a thermometer chart, vertical progress bar, gauge, meter, status indicator, temperature indicator, RAG status bar, or any color-coded level display where fill direction matters.

Discretise your target values so the debounce controller's stability filter can commit. A floating-point target from a noisy producer will never hold for 2 consecutive ticks — the filter blocks every send and the actuator stays dark. Use when wiring a continuous producer signal (confidence score, sensor reading) into a debounced controller, or debugging "I set_target() but _apply() never fires".