Skip to main content
Invoke: /feature-tests AB#1234 · Side effects: commits, never pushes · Reads: context.md, plan.md

​
What it does

  1. Determines the test framework/command from tests.framework / tests.run_command, or auto-detects (package.json → jest/vitest/mocha, pyproject.toml → pytest, *.csproj → dotnet test, go.mod → go test). Polyrepo: does this per touched repo — each can use a different framework.
  2. Writes at least one test per acceptance criterion (a test that would fail if the criterion were unmet), plus at least one test per changed public unit — the traceability map this produces feeds feature-verify and the PR body.
  3. Runs the suite until green, iterating on the code if the implementation is wrong.
If a test can’t pass because the requirement is ambiguous or unachievable as written, this phase stops and surfaces it — it does not delete or dilute a test to force a green run. A passing suite that doesn’t actually check the AC is worse than an honest red.

​
Output

Reports tests added, the AC → test mapping, and the final suite result verbatim (per repo in polyrepo) — including anything still red, since feature-verify and you need the truth, not a green checkmark.