Analyse a feature's source documentation before starting implementation. Reads the documentation from any source (Notion, Confluence, local file, pasted text), explores relevant codebase context, and produces a structured list of questions grouped by target audience (PM, designers, dev lead). Use this skill whenever the user shares requirements, specs, user stories, briefs, or any feature documentation and asks to review it, challenge it, find gaps, or check what's missing before coding. Also trigger when the user provides a Notion/Confluence/Figma link and asks to prepare for implementation, or says things like 'let's check these requirements' or 'what should we clarify before starting'.
Generate a PR description with test plan and record a video demo of the implemented feature using Playwright. Analyses the diff, writes a structured PR body (problem, solution, test plan), generates a volatile Playwright test to record a walkthrough video, and attaches evidence to the PR.
Ship a feature: simplify code, run local CI checks, create/update PR, monitor CI, and run code review. Use when: (1) user invokes /ship, (2) user asks to push or create a PR, (3) agent determines a feature implementation is complete and ready to push.
Drive disciplined TDD implementation of a feature whose acceptance scenarios are already defined and approved (typically BDD/Gherkin .feature files, but any executable acceptance specification works). Use this skill whenever the user asks to start implementing a feature, says things like 'let's build this', 'time to code', 'now write the implementation', or invokes /tdd-implement directly. The skill builds an implementation checklist via TaskCreate from the approved scenarios, runs a per-task loop with opportunistic test-first for non-trivial pure units (transformations, parsers, validators, calculators, branching logic) and a mandatory reflection step to catch missing unit and regression tests, and ends only when all acceptance tests are green. Don't bypass this skill in favour of writing code straight through — the discipline it enforces (no upfront test enumeration, no end-of-feature test back-fill, single-unit cycles, reflection after each task) is exactly what's hard to maintain by hand.