Auto TDD Run

Ready 상태인 phase를 TDD 루프로 실행합니다.

Step 1: Get Ready Phase

ISSUE_ID=$(bd ready --json 2>/dev/null | jq -r '.[0].id // empty')
ISSUE_TITLE=$(bd ready --json 2>/dev/null | jq -r '.[0].title // empty')

if [ -z "$ISSUE_ID" ]; then
    echo "❌ No ready phases. Run 'bd list' to see all issues."
    exit 0
fi

echo "📋 Phase: $ISSUE_TITLE ($ISSUE_ID)"
bd update "$ISSUE_ID" --status in_progress

Step 2: Detect Project Commands

if [ -f "package.json" ]; then
    TEST_CMD="npm test"
    LINT_CMD="npm run lint"
    TYPE_CMD="npm run type-check"
    BUILD_CMD="npm run build"
    E2E_CMD="npm run test:e2e 2>/dev/null || npx playwright test"
fi

Step 3: Execute TDD Phase

Get the issue info and build TDD prompt:

# Re-fetch issue info (variables don't persist between bash calls)
ISSUE_ID=$(bd ready --json 2>/dev/null | jq -r '.[0].id // empty')
ISSUE_TITLE=$(bd ready --json 2>/dev/null | jq -r '.[0].title // empty')

# Detect project commands
if [ -f "package.json" ]; then
    TEST_CMD="npm test"
    LINT_CMD="npm run lint"
    TYPE_CMD="npm run type-check"
    BUILD_CMD="npm run build"
fi

# Build the TDD prompt
RALPH_PROMPT="═══ PHASE: $ISSUE_TITLE ═══
Issue: $ISSUE_ID

CONTEXT RECOVERY:
  bd show $ISSUE_ID
  Read plan file: docs/plans/PLAN_*.md (find matching plan file)

TDD WORKFLOW:
🔴 RED: Write failing tests → Run: $TEST_CMD → Expect FAIL
   ➡ Update plan: Check off RED tasks in plan file

🟢 GREEN: Implement code → Run: $TEST_CMD → Expect PASS
   ➡ Update plan: Check off GREEN tasks in plan file

🔵 REFACTOR:
   Use Task tool with subagent_type='auto-tdd-executor:tdd-refactor' to refactor.
   Then run: $TEST_CMD → Confirm PASS
   ➡ Update plan: Check off REFACTOR tasks in plan file

QUALITY GATE (MANDATORY - DO NOT SKIP):
  ⚠️ MUST run this script (NOT manual commands):
  Find and run: \$HOME/.claude/plugins/*/auto-tdd-marketplace/auto-tdd-executor/*/scripts/quality-gate.sh --phase '$ISSUE_TITLE'
  Or run: find \$HOME/.claude/plugins -name 'quality-gate.sh' -path '*auto-tdd*' -exec {} --phase '$ISSUE_TITLE' \\;

  This script:
  - Runs all tests, lint, type-check, build
  - Auto-updates plan document (QUALITY_GATE + COMPLETE sections)
  - Marks Phase as ✅ in Summary table

  ❌ DO NOT run npm test/lint manually - plan won't update!

VERIFY PLAN UPDATED:
  After quality-gate.sh, check plan file:
  - Phase Summary table shows ✅ (not ⏳)
  - All checkboxes in Phase section are [x]
  If NOT updated → run quality-gate.sh again with --phase flag!

ON QUALITY GATE PASS:
  Run E2E tests:
    Use Task tool with subagent_type='playwright-test-healer'
    to run and fix any E2E failures.

  If E2E pass:
    git add . && git commit -m 'Complete: $ISSUE_TITLE'
    git push origin HEAD
    bd close $ISSUE_ID --reason 'All gates passed'
    <promise>PHASE_COMPLETE</promise>

IF STUCK (30+ iterations):
  bd update $ISSUE_ID --notes 'Blocked: [reason]'
  <promise>PHASE_COMPLETE</promise>

⚠️ CRITICAL:
- NEVER skip quality-gate.sh script
- NEVER use manual npm commands for quality gate
- ALWAYS verify plan document updated before commit"

# Call ralph-loop setup script directly
RALPH_SCRIPT="$HOME/.claude/plugins/cache/claude-plugins-official/ralph-loop/latest/scripts/setup-ralph-loop.sh"

if [ ! -f "$RALPH_SCRIPT" ]; then
    RALPH_SCRIPT="$HOME/.claude/plugins/marketplaces/claude-plugins-official/plugins/ralph-loop/scripts/setup-ralph-loop.sh"
fi

if [ -f "$RALPH_SCRIPT" ]; then
    "$RALPH_SCRIPT" "$RALPH_PROMPT" --max-iterations 50 --completion-promise "PHASE_COMPLETE"
else
    echo "❌ Error: ralph-loop script not found. Please install ralph-loop plugin."
    exit 1
fi

Step 4: Post-Execution

echo ""
echo "═══════════════════════════════════════════════════════════════"
echo "Phase execution completed!"
echo "═══════════════════════════════════════════════════════════════"
bd list
echo ""
echo "Next ready phases:"
bd ready
echo ""
echo "Run '/auto-tdd-run' for next phase"