Phase 3: Verify

Purpose

Confirm implementation matches spec exactly. No bugs, no drift, no missing pieces.

Process

1. Select verification capabilities:

   • Test runners available? (pytest, jest, etc.)
   • Type checkers? (mypy, tsc, etc.)
   • Linters? (eslint, ruff, etc.)
   • MCP servers for validation? (API testing, database state checks)
   • Static analysis tools?
   • Document which tools will be used for verification

2. Automated verification: Run all available checks:

   • All tests pass
   • Type checking passes
   • Linting passes
   • Build succeeds (if applicable)
   • Any project-specific checks (pre-commit hooks, etc.)

   If any fail: Fix and re-run. Do not proceed until green.

3. Spec compliance check: For each item in the spec:

   • Behavior implemented? (trace code path)
   • Edge case handled? (find the code)
   • Test exists for this? (link to test)

   Create verification matrix in docs/scratch//verify.md:

   | Spec Item              | Implemented | Test       | Status |
   |------------------------|-------------|------------|--------|
   | User login             | auth.py:45  | test_auth:12 | ✓    |
   | Invalid password error | auth.py:67  | test_auth:34 | ✓    |
   | Rate limiting          | ???         | ???        | MISSING |
   

4. Scope compliance check:

   • List all files changed (from git diff)
   • Compare to spec's "Files Affected"
   • Flag any changes not in spec
   • Flag any code not traceable to spec requirement

5. Gap handling:

   • If spec item is missing implementation: implement it, return to step 2
   • If implementation exists not in spec: escalate (scope creep detected)
   • If tests missing for spec item: write them, return to step 2

Checkpoint

Present verification matrix to user:

• All automated checks passing
• All spec items traced to code and tests
• No scope creep detected

Escalation

• Spec item cannot be implemented as written
• Discovered requirement that conflicts with existing code
• Verification reveals design flaw requiring spec change