Post-Setup First Cycle

Scenario: The project was just set up (via either spec recipe). The spec documents are filled, cascade passes, scaffold is in place. Now run the first implementation loop.

Prerequisites

  • check_cascade passes all 5 steps
  • docs/PRD.md has ≥1 numbered goal
  • docs/use-cases.md has ≥1 use case in UC-NNN format
  • At least one ADR in docs/adr/

Step 1 — Generate Loop 1 session prompt 

forgecraft_actions({
  action: "generate_session_prompt",
  project_dir: "/path/to/your-project",
  loop: 1
})

The session prompt is written to docs/session-prompts/LOOP-001.md. It contains:

  • Your PRD goals for this loop
  • The first 2–3 use cases
  • Active quality gates for your tags
  • Explicit TDD instructions
  • Loop exit criteria

Step 2 — Open a fresh AI session

Open a new Claude or Copilot session. Paste the content of LOOP-001.md as the first message (or attach the file). Do not carry over prior context — the session prompt is self-contained.

Step 3 — Follow strict TDD

For each behavior in the session prompt:

  1. Write the test first — name it as a specification: it('returns 404 when user not found', ...)
  2. Run the test — it must fail (RED)
  3. Commit the failing test: git commit -m "test(scope): [RED] <behavior>"
  4. Write minimum code to pass the test
  5. Run all tests — all must pass
  6. Commit the implementation: git commit -m "feat(scope): <behavior>"
  7. Refactor if needed, run tests, commit separately

Never collapse RED and GREEN into a single commit. The commit log is evidence of TDD discipline.

Step 4 — Close the cycle

When all Loop 1 behaviors have passing tests:

forgecraft_actions({
  action: "close_cycle",
  project_dir: "/path/to/your-project"
})

close_cycle checks:

  • Cascade still green
  • New tests added (test count increased)
  • No stub implementations in src/
  • Coverage ≥ configured threshold

If blocked, fix the specific issue reported and re-run.

Step 5 — Proceed to Loop 2 

forgecraft_actions({
  action: "generate_session_prompt",
  project_dir: "/path/to/your-project",
  loop: 2
})

Loop 2 session prompt focuses on integration, error handling, and edge cases from your use cases. Repeat Steps 2–4.

Loop 3 and Hardening

Loop 3 focuses on completeness: all use cases covered, all gate checks green, Playwright/pty smoke tests passing. After Loop 3:

forgecraft_actions({
  action: "start_hardening",
  project_dir: "/path/to/your-project"
})

See Post-Hardening for the hardening recipe.