projecta-rrr 1.16.3 → 1.16.4

package/CHANGELOG.md

@@ -4,6 +4,43 @@ All notable changes to RRR will be documented in this file.
 
 Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
+## [1.16.4] - 2026-01-27
+
+### Fixed
+- **Playwright MCP Detection** - Enhanced detection logic now checks multiple locations:
+  - `~/.claude/mcp.json` (primary)
+  - `~/.claude/settings.json` (alternate)
+  - `playwright-mcp` command availability
+  - Running processes via `ps`
+- **QA Engineer Mode Auto-Detection** - UAT now auto-detects test type from SUMMARY.md:
+  - Web UI tests → Recommend AI QA Engineer (Playwright + MiniMax)
+  - API/Backend tests → Recommend Zeroshot
+  - Mixed tests → Offer both options with clear recommendations
+
+### Added
+- **Intuitive Mode Selection** - Testing mode selection now shows:
+  - Auto-detected test type (e.g., "Web UI tests")
+  - Clear explanation of each mode's strengths
+  - Recommended option highlighted
+  - Cost estimates for AI-powered modes
+
+- **run_qa_engineer_batch Step** - New workflow step for AI QA Engineer mode:
+  - Combines Playwright MCP (browser control) + MiniMax (visual AI)
+  - Tests edge cases automatically
+  - Explores navigation paths beyond happy path
+  - ~$0.001 per test (MiniMax vision is 50-100x cheaper than Anthropic API)
+
+- **run_playwright_batch Step** - New workflow step for Playwright-only mode:
+  - Browser automation without visual AI validation
+  - Fallback when MiniMax API key not available
+
+### Changed
+- **Improved Mode Routing** - create_uat_file step now routes to 4 modes:
+  - `qa-engineer` → run_qa_engineer_batch
+  - `zeroshot` → run_zeroshot_batch
+  - `playwright` → run_playwright_batch
+  - `manual` → present_test
+
 ## [1.16.3] - 2026-01-27
 
 ### Changed

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "projecta-rrr",
-  "version": "1.16.3",
+  "version": "1.16.4",
   "description": "A meta-prompting, context engineering and spec-driven development system for Claude Code by Projecta.ai",
   "bin": {
     "projecta-rrr": "bin/install.js"

package/rrr/workflows/verify-work.md

@@ -179,31 +179,122 @@ Check if zeroshot validation is available:
 
 **If ZEROSHOT_AVAILABLE and MODULE_EXISTS:**
 
-Check for Playwright MCP availability:
+Check for Playwright MCP availability (multiple detection methods):
+
+```bash
+# Enhanced Playwright MCP detection - checks multiple locations
+PLAYWRIGHT_STATUS="NO_PLAYWRIGHT_MCP"
+
+# Method 1: Check ~/.claude/mcp.json
+if [ -f "$HOME/.claude/mcp.json" ]; then
+  if grep -qi "playwright" "$HOME/.claude/mcp.json" 2>/dev/null; then
+    PLAYWRIGHT_STATUS="PLAYWRIGHT_MCP_AVAILABLE"
+  fi
+fi
+
+# Method 2: Check ~/.claude/settings.json (alternate location)
+if [ "$PLAYWRIGHT_STATUS" = "NO_PLAYWRIGHT_MCP" ] && [ -f "$HOME/.claude/settings.json" ]; then
+  if grep -qi "playwright" "$HOME/.claude/settings.json" 2>/dev/null; then
+    PLAYWRIGHT_STATUS="PLAYWRIGHT_MCP_AVAILABLE"
+  fi
+fi
+
+# Method 3: Check if playwright-mcp command exists
+if [ "$PLAYWRIGHT_STATUS" = "NO_PLAYWRIGHT_MCP" ]; then
+  if command -v playwright-mcp >/dev/null 2>&1; then
+    PLAYWRIGHT_STATUS="PLAYWRIGHT_MCP_AVAILABLE"
+  fi
+fi
+
+# Method 4: Check MCP servers via ps or other running processes
+if [ "$PLAYWRIGHT_STATUS" = "NO_PLAYWRIGHT_MCP" ]; then
+  if ps aux 2>/dev/null | grep -qi "[p]laywright-mcp"; then
+    PLAYWRIGHT_STATUS="PLAYWRIGHT_MCP_AVAILABLE"
+  fi
+fi
+
+echo "$PLAYWRIGHT_STATUS"
+```
+
+Check for MiniMax API key (enables AI visual validation):
+
+```bash
+[ -n "$MINIMAX_API_KEY" ] && echo "MINIMAX_AVAILABLE" || echo "NO_MINIMAX"
+```
+
+**Determine Recommended Mode:**
+
+Based on what's available, calculate the recommended mode:
+
+| Available | Recommended Mode | Why |
+|-----------|-----------------|-----|
+| Playwright MCP + MiniMax | AI QA Engineer | Best for web UI - browser control + visual validation |
+| MiniMax only | Zeroshot (Screenshot) | Fast, cheap visual validation |
+| Anthropic API only | Zeroshot (Screenshot) | Uses Claude vision |
+| No API keys | Manual | User testing required |
+
+**Auto-detect test type from SUMMARY.md:**
+
 ```bash
-# Check if Playwright MCP is configured
-[ -f "$HOME/.claude/mcp.json" ] && grep -q "playwright" "$HOME/.claude/mcp.json" && echo "PLAYWRIGHT_MCP_AVAILABLE" || echo "NO_PLAYWRIGHT_MCP"
+# Check SUMMARY.md content to determine test type
+SUMMARY_CONTENT=$(cat .planning/phases/*/phases/*-SUMMARY.md 2>/dev/null | head -100)
+
+if echo "$SUMMARY_CONTENT" | grep -qiE "(ui|page|screen|button|form|input|click|navigation|redirect|layout|visual|interface)"; then
+  TEST_TYPE="WEB_UI"
+elif echo "$SUMMARY_CONTENT" | grep -qiE "(api|endpoint|database|backend|voice|sip|telephony|audio|websocket)"; then
+  TEST_TYPE="API_BACKEND"
+else
+  TEST_TYPE="MIXED"
+fi
+echo "$TEST_TYPE"
+```
+
+**Mode Selection UX:**
+
+```
+╔══════════════════════════════════════════════════════════════════════╗
+║  UAT Testing Mode (auto-detected: Web UI tests)                      ║
+╠══════════════════════════════════════════════════════════════════════╣
+║                                                                      ║
+║  Recommended:                                                       ║
+║  ────────────────────                                               ║
+║  ⭐ AI QA Engineer (Recommended for web UI)                          ║
+║     • Browser automation + visual AI validation                     ║
+║     • Explores edge cases, tests like a real user                   ║
+║     • Cost: ~$0.001/test (MiniMax vision)                           ║
+║                                                                      ║
+║  Alternative:                                                       ║
+║  ────────────                                                       ║
+║  • Zeroshot (Screenshot) — Fast, fully automated                     ║
+║  • Manual — You test each feature and report results                 ║
+║                                                                      ║
+╚══════════════════════════════════════════════════════════════════════╝
 ```
 
 Use AskUserQuestion:
 - header: "Testing Mode"
-- question: "How would you like to run UAT?"
+- question: "How would you like to run UAT? (Recommended: AI QA Engineer for web UI)"
 - options:
-  {If PLAYWRIGHT_MCP_AVAILABLE:}
-  - "AI QA Engineer (Recommended)" — Claude acts as a QA engineer, controlling the browser interactively. Explores like a real user, tests edge cases.
+  {If PLAYWRIGHT_MCP_AVAILABLE and MINIMAX_AVAILABLE:}
+  - "AI QA Engineer (Recommended)" — Claude acts as a QA engineer, controlling the browser interactively with visual validation. Tests edge cases automatically. Best for web UI.
   - "Zeroshot (Screenshot)" — Claude validates UI using screenshots. Faster, less thorough.
   - "Manual" — You'll test each feature and report results. More control, slower.
-  {Else:}
+  {ElseIf PLAYWRIGHT_MCP_AVAILABLE and TEST_TYPE = "WEB_UI":}
+  - "Playwright MCP (Interactive)" — Use browser automation for testing. No visual AI validation.
+  - "Manual" — You'll test each feature and report results.
+  {ElseIf MINIMAX_AVAILABLE or ANTHROPIC_API_KEY:}
   - "Zeroshot (Automated) (Recommended)" — Claude validates UI autonomously using screenshots. Faster, no input needed per test.
   - "Manual" — You'll test each feature and report results. More control, slower.
+  {Else:}
+  - "Manual" — You'll test each feature and report results.
 
-Store result in `TESTING_MODE` (qa-engineer, zeroshot, or manual).
+Store result in `TESTING_MODE` (qa-engineer, zeroshot, playwright, or manual).
 
 **If MANUAL_ONLY or MODULE_MISSING:**
 
 ```
 UAT will run in manual mode.
-(Set ANTHROPIC_API_KEY to enable automated zeroshot testing)
+(Set ANTHROPIC_API_KEY to enable automated testing)
 ```
 
 Set `TESTING_MODE=manual`.
@@ -353,7 +444,9 @@ skipped: 0
 Write to `{phase_dir}/{phase}-UAT.md` (where phase_dir is from get_phase_dir)
 
 **Route based on testing mode:**
+- If `TESTING_MODE=qa-engineer` → Go to `run_qa_engineer_batch`
 - If `TESTING_MODE=zeroshot` → Go to `run_zeroshot_batch`
+- If `TESTING_MODE=playwright` → Go to `run_playwright_batch`
 - If `TESTING_MODE=manual` → Go to `present_test`
 </step>
 
@@ -444,6 +537,128 @@ Falling back to manual mode for remaining tests.
 Set `TESTING_MODE=manual` and continue from current test.
 </step>
 
+<step name="run_qa_engineer_batch">
+**Run AI QA Engineer mode for all tests:**
+
+This mode combines Playwright MCP (browser control) with MiniMax (visual AI) for comprehensive exploratory testing.
+
+```
+╔══════════════════════════════════════════════════════════════╗
+║  AI QA ENGINEER: Autonomous Testing                          ║
+╚══════════════════════════════════════════════════════════════╝
+
+Running {N} tests with browser automation + AI visual validation...
+```
+
+**For each test in the UAT file:**
+
+1. Extract test name and expected behavior
+2. Determine the URL to test
+3. Run QA Engineer validation:
+   ```javascript
+   const { zeroshotValidate } = require('~/.claude/rrr/lib/uat/zeroshot-validator.js');
+
+   const result = await zeroshotValidate({
+     url: testUrl,
+     task: `Verify: ${expectedBehavior}`,
+     mode: 'qa-engineer',        // QA engineer persona
+     interactive: true,          // Use Playwright MCP for browser control
+     timeout: 60000              // Longer timeout for exploratory testing
+   });
+   ```
+
+4. Process result:
+   - If `result.passed === true`:
+     - Update test result to "pass"
+     - Log: `✓ Test {N}: {name}`
+   - If `result.passed === false`:
+     - Update test result to "issue"
+     - Capture observations and reasoning
+     - Add to Gaps section
+     - Log: `✗ Test {N}: {name} - {brief reason}`
+
+5. Update UAT.md after each test (checkpoint)
+
+6. Display progress:
+   ```
+   Progress: [████████░░] 8/10 tests complete
+   ✓ Test 1: Login page loads with form fields
+   ✓ Test 2: Form validation on submit
+   ✗ Test 3: Password reset email not sent (edge case found)
+   ...
+   ```
+
+**QA Engineer specific behaviors:**
+- Tests edge cases automatically (empty fields, long inputs, special chars)
+- Explores navigation paths beyond the happy path
+- Screenshots any issues found
+- Validates visual correctness with MiniMax vision
+
+**After all tests complete:**
+
+Update frontmatter status to "complete".
+
+**Display summary:**
+
+```
+╔══════════════════════════════════════════════════════════════╗
+║  AI QA ENGINEER: Complete                                     ║
+╚══════════════════════════════════════════════════════════════╝
+
+Results:
+  ✓ Passed:  {N}
+  ✗ Issues:  {N}
+  ⏭ Skipped: {N}
+
+Exploratory Testing Notes:
+  - {N} edge cases tested automatically
+  - {N} navigation paths explored
+
+{If issues > 0:}
+Issues Found:
+  1. Test 3: Password reset - Edge case: special chars in email cause error
+  2. Test 7: Form submission - Visual: error message clipped on mobile
+```
+
+Go to `complete_session`.
+
+**On error (Playwright MCP not available, timeout):**
+
+```
+⚠ QA Engineer mode error: {error}
+Falling back to Zeroshot mode...
+```
+
+Set `TESTING_MODE=zeroshot` and continue to `run_zeroshot_batch`.
+</step>
+
+<step name="run_playwright_batch">
+**Run Playwright-only mode (no visual AI):**
+
+Use Playwright MCP for browser automation without visual AI validation.
+
+```
+╔══════════════════════════════════════════════════════════════╗
+║  PLAYWRIGHT UAT: Browser Automation                           ║
+╚══════════════════════════════════════════════════════════════╝
+
+Running {N} tests with Playwright browser automation...
+```
+
+**For each test:**
+
+1. Navigate to URL using Playwright MCP
+2. Perform interactions described in test
+3. Take screenshot at key points
+4. Report findings (manual AI analysis not used)
+
+**After all tests:**
+
+Display summary and prompt for manual verification of screenshots.
+
+Go to `complete_session`.
+</step>
+
 <step name="present_test">
 **Present current test to user (Manual Mode):**