Review the diff since a fixed point (commit, branch, tag, merge-base) along two independent axes — Standards (does it follow the repo's coding standards and avoid known smells?) and Spec (does it match what the originating issue/spec asked for?). Runs one isolated sub-agent per axis and reports them side by side…
Actively build and sharpen a project's domain model as you design — challenge fuzzy terms, invent edge-case scenarios, and write the glossary and architectural decisions down the moment they crystallise. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision…
Prepare a repository for agent work — install the deterministic verification layer (formatter, linter, types, tests, git hooks, CI) without breaking what already exists, and record the canonical commands and prohibitions in the repo's agent-context file (AGENTS.md or CLAUDE.md). Use whenever the user wants to set up a…
Execute one already-written specification (a .md file, PRD, or GitHub issue) into one reviewable PR, with closed scope and attached evidence. Delegates the test loop to tdd and the review to code-review, and reads the canonical commands from the repo's agent-context file (AGENTS.md or CLAUDE.md). Use whenever the user…
Test-driven development — the red-green loop, done so the tests are worth keeping. Use when building a feature or fixing a bug test-first, when the user mentions red-green-refactor, or when spec-execution reaches an agreed test seam.
Turn the current conversation — ideally the decision log from a grill session — into a spec (a PRD) and publish it to the project's issue tracker, or to a local file if there is none. No interview; this is pure synthesis of what has already been decided. Use when the user says "write the spec", "turn this into a…
Break a spec, plan, or settled conversation into a set of tracer-bullet tickets, each declaring the tickets that block it, published to the configured tracker — native blocking links on a real tracker, or one file per ticket locally. Use when a spec is too big for one PR and needs slicing into dependency-ordered…
A relentless interview that stress-tests a plan, design, or decision before any code is written — one question at a time, each with a recommended answer, walking the full decision tree until nothing is left ambiguous. It challenges choices that contradict your own stated principles, scores options against your…
The specflow router — map what the user is trying to do to the right specflow skill and the right point in the flow. Use when the user explicitly signals they do not know which step to take: "which specflow skill do I use", "where do I start with this", "how do I begin", "what's next in the flow", "I have a spec/idea…
Compact the current session into an ephemeral handoff document so another session or tool can pick the work up once and continue. Records session state and git state, references durable artifacts by path instead of duplicating them, and suggests which skills to run next. Use when you have to stop mid-work and hand…