Generates JUnit regression tests that lock in current behavior before a refactor, capturing observed outputs as assertions so that any behavioral change trips a test. Use before large refactors, when inheriting untested legacy Java, or when the spec is "whatever it does now."

Updates broken pytest tests after intentional code changes — triaging assertion failures from mock-coupling failures from genuine regressions, using Python's introspection to automate where safe. Use when a refactor or API change leaves a pile of failing tests and you need to decide update vs. fix vs. delete.

Generates hardened, multi-stage Dockerfiles with non-root users, minimal base images, and a .dockerignore, after auto-detecting the application stack. Use when containerizing an application for the first time, when the user asks for a Dockerfile, when migrating from a VM deployment, or when an existing Dockerfile runs as root, uses a fat base image, or leaks build tooling into the runtime layer.

Executes refactorings — extract method, inline, rename, move — in small, behavior-preserving steps with a test between each. Use when the user wants to restructure working code, when cleaning up after a feature lands, or when a smell has been identified and needs fixing.

Derives executable test cases directly from requirements (user stories, acceptance criteria, specs) by extracting testable conditions, enumerating equivalence classes and boundaries, and producing a traceability map from each test back to its source requirement. Use when building acceptance tests from a spec, when checking whether requirements are covered by existing tests, when translating Gherkin or plain-English criteria into code, or when proving coverage for compliance.

Extracts a TLA+ specification from concurrent or distributed code, modeling the state machine, actions, and fairness conditions for model checking with TLC. Use when verifying concurrency properties of production code, when designing a protocol and wanting to check it before implementation, or when the user has a race condition and needs to prove the fix.

Verifies that a refactoring or transformation preserved observable behavior by comparing before and after execution, differential testing, or I/O capture. Use after a refactoring, after automated code transformation, before merging a structural PR, or whenever the claim is that two code versions do the same thing.

Translates natural-language or pseudocode descriptions of concurrent and distributed systems into TLA+ specifications ready for the TLC model checker. Identifies state variables, actions, type invariants, safety properties, and liveness properties from the description. Use when formalizing a protocol, when the user describes a distributed algorithm to verify, when designing a consensus or locking scheme, or when starting formal verification of a concurrent system.

Applies the SZZ algorithm to VCS history to identify which commits introduced bugs by correlating bug-fix commits with earlier changes. Use when mining a repository for bug-introducing commits, when building a defect-prediction dataset, or when the user asks which commit introduced a given fixed bug.

Translates an entire module or package between languages, handling imports, file layout, visibility, and cross-function dependencies that single-function translation misses. Use when porting a library, when a migration spans multiple files, or when the user hands you a directory and a target language.

Migrates a Spring MVC application to Spring Boot, converting XML config to auto-configuration, restructuring the project, and replacing container deployment with embedded. Use when modernizing a legacy Spring app, when moving off a standalone servlet container, or when the user has web.xml and wants application.yml.

Vue 3 Composition API, script setup macros, reactivity system, and built-in components. Use when writing Vue SFCs, defineProps/defineEmits/defineModel, watchers, or using Transition/Teleport/Suspense/KeepAlive.

Node.js package manager with strict dependency resolution. Use when running pnpm specific commands, configuring workspaces via pnpm-workspace.yaml, or managing dependencies with catalogs, patches, overrides, config dependencies, or the global virtual store.

Create, validate, install, fold, or troubleshoot Codex subagent role TOML, agents-table config, discoverability wiring, and duplicate-role merges. Use when a user asks for a Codex agent role, reviewer agent, role config, TOML role file, subagent setup, or overlapping agents to merge.

Outsider-perspective end-to-end review of a plan, PR, or code change. First questions intent and whether a simpler/more elegant approach would achieve the same goal, then traces the actual code path (not just the diff) to verify the change does what it claims. Output is concise, actionable, and every call carries its rationale. Trigger on /scrutinize and proactively whenever the user asks to review, audit, sanity-check, or get a second opinion on a plan, PR, diff, design doc, or proposed code change.

Use when working with docusaurus.config.js/ts files — validate the config structure, set required fields (title/url/baseUrl), fix URL and baseUrl formats, add presets and plugins, configure navbar/footer/theme settings, and move unknown keys into customFields. Triggers on tasks involving docusaurus.config.js, docusaurus.config.ts, site configuration, URL formats, presets, or theme config.

Generate and deploy production-grade websites from a natural language description. Use when asked to "vibe a site", "vibe code this", "build me a website and publish it", "create a landing page and deploy it", "make a portfolio site and put it online", "generate a site and ship it", or any request combining website creation with deployment. Produces a live URL via here.now.

Build a narrated demo video from a screen recording, a talk track script with act timelines, branded intro/outro slides, background music, and AI-generated voiceover via ElevenLabs. Use when user asks to create a "demo video", "product demo", "narrated screencast", "talk track video", or "voiceover video".

Stress-test a plan, design, or idea against the project's domain language (CONTEXT.md) and recorded decisions (docs/adr/). One question at a time, walks the decision tree, captures crystallized terminology and decisions inline. THIS IS THE DEFAULT GRILLING SKILL — use for any interrogation in a real project, even if CONTEXT.md or docs/adr/ don't exist yet (they get created lazily on first term resolution / first ADR-worthy decision). Triggers: /grill-with-docs, "grill me", "grill this", "interrogate this plan", "stress-test this design", "walk me through the decisions", "challenge this". Use grill-me ONLY for pure green-field thinking with no codebase at all — rare.

Read an existing PRD (typically issues/prd.md, written by write-a-prd) and break it into vertical-slice tracer-bullet issues at issues/NNN-short-title.md, tagged HITL (needs human) or AFK (autonomous). Step 2 of the build pipeline (write-a-prd → prd-to-issues → work-issues). Use when a PRD exists and needs to become work tickets. NOT for writing the PRD itself (use write-a-prd). Triggers: /prd-to-issues, "break this PRD into issues", "turn this spec into tickets".