React Feature Kit — Testing skill for Claude Code

React Feature Kit

Spec-driven development for React in [Claude Code](https://claude.com/claude-code). Three plugins: a workflow that builds a feature from a spec, and two structure plugins that keep files where the project says they go.

The problem

One long chat produces one big diff and nothing to check it against. The spec was in your head, the plan was in the model's, and by the time you review there is no document that says what "done" meant.

The kit splits that chat into stages. Each stage writes a file, and the next stage starts a fresh session and reads it.

/spec       .planning//SPEC.md       what the user sees, acceptance criteria with ids
/plan       PLAN.md + contract.md          modules in build order, each citing the criteria it serves
/implement  code                           reads the spec and the plan, builds in order
/verify     VERIFY.md                      every criterion: pass, fail or unverifiable, with evidence
/review     REVIEW.md                      code rules, needless complexity, wasted work
/refine     code                           applies what verify and review found, one item at a time

You read and fix `SPEC.md` before a single module exists. Request and response shapes come from your OpenAPI file, cut down to the endpoints the spec names. Figma frames are read by subagents, so their payload never lands in your session.

Install

claude plugin marketplace add vadimgaidai/react-feature-kit
claude plugin install react-feature-workflow@vadimgaidai --scope project

Then one structure plugin, if the project has a structure to enforce:

claude plugin install feature-sliced-design@vadimgaidai --scope project   # FSD projects
claude plugin install feature-folders@vadimgaidai --scope project         # everything else

`--scope project` matters. Without it the plugins load in every repository you open. Details in [docs/GETTING-STARTED.md](./docs/GETTING-STARTED.md).

What to type