Review Million Dollar App

Context

• Upstream:
  • /build-million-dollar-app scaffolds project
  • /plan-million-dollar-app creates plan
  • /implement-million-dollar-app implements features
  • /verify-million-dollar-app collects evidence and submits for review
• Downstream:
  • APPROVE: Terminal success - status becomes built
  • DENY: Loops back - status becomes scaffolded_available with cumulative feedback

Scope

• Input: Idea slug from review queue OR shows all pending reviews
• Ready State: Reviews ideas with status pending_review
• APPROVE Outcome: Status -> built (terminal success)
• DENY Outcome: Status -> scaffolded_available + cumulative feedback

This skill does NOT:

• Auto-approve (NEVER - only human can approve)
• Modify implementation code
• Deploy the application

This skill CAN:

• Auto-deny iterations 1-5 by default (configurable with --auto-deny=N or disabled with --no-auto-deny)
• Spawn verification agents to automatically discover issues during auto-deny

Output

• On APPROVE: Tracking updated to status: built
• On DENY:
  • Tracking updated to status: scaffolded_available
  • Feedback summary appended to tracking.ideas[slug].denial_feedback[]
  • Detailed feedback appended to [project]/.planning/feedback.md

AUTO-DENY FEATURE

Default Behavior

By default, the first 5 review iterations are automatically denied. This allows the system to iterate quickly without requiring human attention for early-stage issues.

Why Auto-Deny?

1. Early iterations have obvious issues - First few attempts often have configuration errors, missing dependencies, or basic functionality gaps
2. Human time is precious - Don't waste human review on apps that can't even start
3. Faster iteration cycles - The system can run 24/7, finding and fixing issues automatically
4. Cumulative learning - Each auto-deny adds structured feedback that accumulates

Auto-Deny Threshold Logic

review_count < auto_deny_threshold  -->  AUTO-DENY (agent finds issues)
review_count >= auto_deny_threshold -->  HUMAN REVIEW required

Default threshold: 5

Configuring Auto-Deny

# Default: auto-deny first 5 iterations
/review-million-dollar-app simple-todo-app

# Custom threshold: auto-deny first 3 iterations
/review-million-dollar-app simple-todo-app --auto-deny=3

# Disable auto-deny: always require human decision
/review-million-dollar-app simple-todo-app --no-auto-deny

# Auto-deny first 10 iterations
/review-million-dollar-app simple-todo-app --auto-deny=10

Auto-Deny Verification Agent

When auto-denying, spawn a verification agent to discover issues:

Task({
  subagent_type: "general-purpose",
  description: "Auto-deny verification agent",
  prompt: `
    You are verifying the app at: ${projectPath} for auto-deny review #${reviewCount + 1}

    Run these checks and report ALL issues found:

    1. **Database & Infrastructure**
       - Check if docker-compose.yml exists and is valid
       - Verify .env has real values (not placeholders)
       - Check if database can be started

    2. **Build Verification**
       - Run: pnpm install (check for errors)
       - Run: pnpm build (capture any errors)
       - Check for TypeScript/compilation errors

    3. **Runtime Verification**
       - Start the dev server: pnpm dev
       - Wait 10 seconds for startup
       - Check for console errors
       - Try to fetch http://localhost:3000
       - Check for 404s or startup failures

    4. **Test Verification**
       - Run: pnpm test (if tests exist)
       - Report pass/fail counts
       - Note any failing tests

    5. **Code Quality**
       - Check for TODO comments in critical paths
       - Look for placeholder values
       - Verify auth configuration

    Return JSON:
    {
      "status": "issues_found" | "no_issues",
      "issues": [
        {
          "category": "database|build|runtime|tests|code_quality",
          "severity": "critical|major|minor",
          "description": "What's wrong",
          "file": "path/to/file.ts:line" (if applicable),
          "fix_hint": "How to fix it"
        }
      ],
      "summary": "One-line summary of main problems",
      "can_app_start": true|false,
      "tests_passing": "X/Y" | "no tests" | "tests failed"
    }
  `
})

ARCHITECTURE: REVIEW INTERFACE

/review-million-dollar-app (Interface)
    |
    |---> [Step 0] Parse Arguments
    |     - Extract idea-slug
    |     - Determine auto-deny threshold (default: 5)
    |     - Check for --no-auto-deny flag
    |
    |---> [Step 1] Show Review Queue
    |     - List all pending_review items
    |     - Show submission date and review count
    |     - Indicate which are in auto-deny range
    |
    |---> [Step 2] Check Auto-Deny Eligibility
    |     - IF review_count < threshold AND auto-deny enabled:
    |           --> [Step 2a] Auto-Deny Flow
    |     - ELSE:
    |           --> [Step 2b] Human Review Flow
    |
    |---> [Step 2a] Auto-Deny Flow
    |     - Spawn verification agent
    |     - Collect discovered issues
    |     - Generate structured feedback
    |     - Process as DENY (no human input needed)
    |     - Display results to user
    |
    |---> [Step 2b] Human Review Flow
    |     - Present Evidence
    |     - Display test results, screenshots, etc.
    |     - Show prior feedback
    |     - Wait for human decision (APPROVE/DENY)
    |
    |---> [Step 3] Process Decision
          - APPROVE: Update to 'built' (human only)
          - DENY: Update to 'scaffolded_available' + write feedback

ORCHESTRATION WORKFLOW

Step 0: Parse Arguments

// Parse command line arguments
const args = parseArguments(input);
const slug = args.slug;  // May be null (show queue)
let autoDenyThreshold = 5;  // Default

if (args.flags['no-auto-deny']) {
  autoDenyThreshold = 0;  // Disabled
} else if (args.flags['auto-deny']) {
  // --auto-deny or --auto-deny=N
  autoDenyThreshold = args.flags['auto-deny'] === true ? 5 : parseInt(args.flags['auto-deny']);
}

Step 1: Show Review Queue

Starting /review-million-dollar-app...
----------------------------------------------------

Auto-Deny: ENABLED (threshold: 5)

REVIEW QUEUE
============

| # | Idea                    | Submitted  | Reviews | Mode        | Last Denial |
|---|-------------------------|------------|---------|-------------|-------------|
| 1 | ai-code-review-tools    | 2026-01-15 | 1       | AUTO-DENY   | -           |
| 2 | smart-inventory-tracker | 2026-01-16 | 5       | HUMAN       | 2026-01-14  |
| 3 | fitness-meal-planner    | 2026-01-17 | 3       | AUTO-DENY   | 2026-01-16  |

Legend:
- AUTO-DENY: Review count < 5, will be automatically verified and denied
- HUMAN: Review count >= 5, requires human APPROVE/DENY decision

Enter idea slug to review (or number):

Step 2: Present Evidence

When user selects an idea:

REVIEWING: ai-code-review-tools
===============================

## Project Overview
- Location: ~/Developer/million-dollar-app-ideas/ai-code-review-tools
- Type: webapp (Next.js)
- Review attempt: #1

## Test Results
+---------------------+-------------------+
| Category            | Result            |
+---------------------+-------------------+
| Unit Tests          | 45/45 passed      |
| Integration Tests   | 12/12 passed      |
| E2E Tests           | 8/8 passed        |
| Coverage            | 87% lines         |
+---------------------+-------------------+

## Screenshots
1. [login.png] - Login page with GitHub OAuth
2. [dashboard.png] - Main dashboard with repo list
3. [repo-detail.png] - Repository detail view
4. [review-results.png] - AI code review results
5. [settings.png] - User settings page

View screenshots at: evidence/*.png

## Demo
URL: http://localhost:3000
(Start server with: cd [project] && pnpm dev)

## Market Validation
+--------------------------------+----------+---------+
| Claim                          | Verified | Current |
+--------------------------------+----------+---------+
| Market size $X billion         | Yes      | Yes     |
| Competitor Y revenue $Z/month  | Yes      | Yes     |
| TAM/SAM analysis               | Yes      | Yes     |
+--------------------------------+----------+---------+

Differentiation: AI-powered approach provides 10x faster reviews
Revenue Path: Clear (SaaS subscription model)

## Implementation Summary
- Phases completed: 4/4
- Tasks completed: 23/23
- Files created: 47
- Files modified: 12

${hasPriorFeedback ? `
## PRIOR FEEDBACK (Review #${reviewCount - 1})
This idea was previously DENIED. Review the prior feedback:

${priorFeedbackSummary}

The plan should have addressed this feedback. Verify it was incorporated.
` : ''}

----------------------------------------------------

DECISION REQUIRED

[A] APPROVE - Mark as 'built' ($1M ready!)
[D] DENY - Return to planning with feedback

Enter decision (A/D):

Step 2a: Auto-Deny Flow (when review_count < threshold)

This step REPLACES steps 2 and 3 when auto-deny is active.

// Check if auto-deny applies
const reviewCount = tracking.ideas[slug].review_count || 0;
const nextReviewNum = reviewCount + 1;

if (nextReviewNum <= autoDenyThreshold && autoDenyThreshold > 0) {
  console.log(`
AUTO-DENY MODE ACTIVE
=====================
Review iteration: #${nextReviewNum} of ${autoDenyThreshold} (auto-deny threshold)
Mode: Automatic verification and denial

Spawning verification agent...
  `);

  // Spawn auto-deny verification agent
  const verificationResult = await Task({
    subagent_type: "general-purpose",
    description: "Auto-deny verification agent",
    prompt: `
      You are verifying the app at: ${projectPath} for auto-deny review #${nextReviewNum}

      Run these checks and report ALL issues found:

      1. **Database & Infrastructure**
         - Check if docker-compose.yml exists and is valid
         - Verify .env has real values (not placeholders like "your-secret-here")
         - Check if scripts/setup.sh exists and is executable

      2. **Build Verification**
         - Run: pnpm install (check for errors)
         - Run: pnpm build (capture any build/TypeScript errors)

      3. **Runtime Verification**
         - Start database: docker compose up -d (if docker-compose.yml exists)
         - Start dev server: pnpm dev (in background)
         - Wait 15 seconds for startup
         - Check for console errors (look for "Error", "error:", exceptions)
         - Try to curl http://localhost:3000
         - Check for 404s, 500s, or startup failures
         - Kill the dev server when done

      4. **Test Verification**
         - Run: pnpm test (if tests exist)
         - Report pass/fail counts
         - Note any failing tests

      5. **Code Quality**
         - Grep for TODO, FIXME, PLACEHOLDER in src/
         - Look for hardcoded secrets or placeholder values
         - Check .env.example vs .env for missing variables

      Return JSON (MUST be valid JSON):
      {
        "status": "issues_found" | "no_issues",
        "issues": [
          {
            "category": "database|build|runtime|tests|code_quality",
            "severity": "critical|major|minor",
            "description": "What is wrong",
            "file": "path/to/file.ts:line",
            "fix_hint": "How to fix it"
          }
        ],
        "summary": "One-line summary of main problems",
        "can_app_start": true|false,
        "tests_passing": "X/Y" | "no tests" | "tests failed to run"
      }
    `
  });

  // Generate feedback from verification results
  const feedback = generateFeedbackFromVerification(verificationResult);

  // Process as DENY (same as Step 4 but with auto-generated feedback)
  processDeny(slug, feedback, `auto-deny-review-${nextReviewNum}`);

  console.log(`
AUTO-DENIED (Review #${nextReviewNum})
=======================

${slug} has been automatically denied.

Tracking Updates:
- Status: scaffolded_available
- Review count: ${nextReviewNum}
- Feedback: Written to .planning/feedback.md

Auto-Deny Status:
- Iterations remaining: ${autoDenyThreshold - nextReviewNum} more auto-denies before human review
- Next review (#${nextReviewNum + 1}) will ${nextReviewNum + 1 <= autoDenyThreshold ? 'also be auto-denied' : 'REQUIRE human review'}

Next Steps:
1. /plan-million-dollar-app will read feedback and create fix plan
2. /implement-million-dollar-app will apply fixes
3. /verify-million-dollar-app will re-verify
4. /review-million-dollar-app will review again
  `);

  return;  // Exit early, skip human review flow
}

// If we reach here, proceed to human review (Step 2b)

Step 3: Process APPROVE (Human Only)

if (decision === 'APPROVE') {
  // Acquire lock
  const tracking = acquireLock('./ideas/_tracking.json');
  const now = new Date().toISOString();

  // Update status to built
  tracking.ideas[slug].status = 'built';  // Terminal success
  tracking.ideas[slug].current_phase = 'complete';
  tracking.ideas[slug].approved_at = now;

  // Remove from review queue
  tracking.review_queue = tracking.review_queue.filter(r => r.slug !== slug);

  tracking.ideas[slug].history.push({
    timestamp: now,
    action: 'approved',
    details: 'Human review APPROVED. Status: built. $1M ready!'
  });

  // Update meta counts
  updateStatusCounts(tracking);
  tracking.meta.last_updated = now;

  // Release lock
  releaseLock(tracking, './ideas/_tracking.json');

  console.log(`

APPROVED!
=========

${slug} has been marked as 'built'.

This app is now $1M ready!

Status: built (terminal success)

Congratulations!
  `);
}

Step 4: Process DENY (with Cumulative Feedback)

CRITICAL: Feedback is APPENDED, not replaced. Both locations are updated.

if (decision === 'DENY') {
  // Prompt for feedback
  const feedback = await promptForFeedback();

  // Acquire lock
  const tracking = acquireLock('./ideas/_tracking.json');
  const now = new Date().toISOString();

  // Increment review count
  tracking.ideas[slug].review_count = (tracking.ideas[slug].review_count || 0) + 1;
  const reviewNum = tracking.ideas[slug].review_count;

  // 1. APPEND feedback to tracking (summary)
  if (!tracking.ideas[slug].denial_feedback) {
    tracking.ideas[slug].denial_feedback = [];
  }
  tracking.ideas[slug].denial_feedback.push({
    review_number: reviewNum,
    timestamp: now,
    summary: feedback.summary
  });

  // 2. APPEND detailed feedback to .planning/feedback.md
  const feedbackPath = `${projectPath}/.planning/feedback.md`;
  const existingFeedback = fileExists(feedbackPath) ? read(feedbackPath) : getInitialFeedbackTemplate();

  const newFeedbackSection = generateFeedbackSection(reviewNum, now, feedback);
  const updatedFeedback = insertFeedbackAtTop(existingFeedback, newFeedbackSection);

  write(feedbackPath, updatedFeedback);

  // 3. Update status to scaffolded_available (loops back to planning)
  tracking.ideas[slug].status = 'scaffolded_available';
  tracking.ideas[slug].current_phase = 'awaiting_replan';

  // 4. Remove from review queue
  tracking.review_queue = tracking.review_queue.filter(r => r.slug !== slug);

  tracking.ideas[slug].history.push({
    timestamp: now,
    action: 'denied',
    details: `Human review DENIED (#${reviewNum}). Status: scaffolded_available. Feedback written to .planning/feedback.md`
  });

  // Update meta counts
  updateStatusCounts(tracking);
  tracking.meta.last_updated = now;

  // Release lock
  releaseLock(tracking, './ideas/_tracking.json');
}

FEEDBACK COLLECTION INTERFACE

When DENY is selected, collect structured feedback:

DENIAL FEEDBACK
===============

You are denying review #${reviewNum} for: ${slug}

Please provide feedback in each section:

## Issues Found
List the problems discovered during review:
> 1. Auth not working - OAuth callback fails
> 2. Missing error handling in checkout flow
> 3. Tests incomplete for payment edge cases

## Required Changes
What MUST be fixed before next review:
> 1. Fix OAuth callback URL configuration
> 2. Add try/catch and user-friendly errors in checkout
> 3. Add unit tests for payment failure scenarios

## Evidence Gaps (optional)
What evidence was missing or insufficient:
> 1. Market validation sources from 2023, need 2025/2026 data
> 2. No demo video provided

## Additional Notes (optional)
Any other feedback:
> Consider adding loading states to improve UX

----------------------------------------------------

Review feedback:
- Issues: 3
- Required changes: 3
- Evidence gaps: 2

Confirm DENY with this feedback? (Y/N):

FEEDBACK FILE FORMAT

The .planning/feedback.md file is CUMULATIVE - new feedback is PREPENDED:

# Review Feedback

> **CRITICAL**: The /plan-app agent MUST read this entire file before creating a new plan.
> All previous feedback must be addressed in the new plan.

---

## Review #2 - 2026-01-17

**Status**: DENIED

### Issues Found
1. Checkout flow missing error handling for payment failures
2. Unit tests only cover 60% of critical paths
3. No loading states on dashboard

### Required Changes
- Add try/catch and user-friendly error messages in checkout
- Add tests for payment edge cases
- Implement skeleton loaders for dashboard widgets

### Evidence Gaps
- Market validation sources were from 2023, need current data
- No demo video provided

---

## Review #1 - 2026-01-15

**Status**: DENIED

### Issues Found
1. Authentication not working
2. Database migrations incomplete

### Required Changes
- Fix OAuth callback handling
- Run and verify all migrations

---

COMMUNICATION PROTOCOL

Complete Review Flow

User: /review-million-dollar-app

Agent:
Starting /review-million-dollar-app...
----------------------------------------------------

REVIEW QUEUE
============

| # | Idea                    | Submitted  | Reviews | Last Denial |
|---|-------------------------|------------|---------|-------------|
| 1 | ai-code-review-tools    | 2026-01-15 | 1       | -           |

Enter idea slug to review (or 1 for first item): 1

REVIEWING: ai-code-review-tools
===============================

[Evidence presentation as shown above]

----------------------------------------------------

DECISION REQUIRED

[A] APPROVE - Mark as 'built' ($1M ready!)
[D] DENY - Return to planning with feedback

Enter decision (A/D): A

----------------------------------------------------

APPROVED!
=========

ai-code-review-tools has been marked as 'built'.

This app is now $1M ready!

Tracking:
- Status: built
- Review queue: Removed

Congratulations!

Deny Flow with Feedback

Enter decision (A/D): D

DENIAL FEEDBACK
===============

## Issues Found
> 1. Login redirect loop on mobile
> 2. Dashboard loads slowly (5+ seconds)

## Required Changes
> 1. Fix mobile OAuth redirect handling
> 2. Add pagination to dashboard data fetching

## Evidence Gaps
> (none)

Confirm DENY with this feedback? (Y/N): Y

----------------------------------------------------

DENIED
======

ai-code-review-tools has been returned to planning phase.

Tracking:
- Status: scaffolded_available
- Review count: 2
- Feedback: Written to .planning/feedback.md

Next Steps:
1. /plan-million-dollar-app will read .planning/feedback.md
2. New plan MUST address all feedback
3. After re-implementation: /verify-million-dollar-app
4. Then back to /review-million-dollar-app

The feedback loop continues until APPROVED or ABANDONED.

TRACKING SYSTEM

File: ./ideas/_tracking.json

This skill interacts with the tracking system based on human decision:

On APPROVE:

{
  "ideas": {
    "idea-slug": {
      "status": "built",
      "current_phase": "complete",
      "approved_at": "ISO-8601"
    }
  },
  "review_queue": []  // Item removed from queue
}

On DENY:

{
  "ideas": {
    "idea-slug": {
      "status": "scaffolded_available",
      "current_phase": "awaiting_replan",
      "review_count": 2,  // Incremented
      "denial_feedback": [
        {
          "review_number": 1,
          "timestamp": "ISO-8601",
          "summary": "Previous feedback summary"
        },
        {
          "review_number": 2,
          "timestamp": "ISO-8601",
          "summary": "New feedback summary appended"
        }
      ]
    }
  },
  "review_queue": []  // Item removed from queue
}

Locking Protocol

1. Acquire: Check lock is null or expired, then set with 5-minute expiry
2. Operate: Make changes to tracking file
3. Release: Set lock to null
4. Auto-expire: Locks older than 5 minutes are considered stale

Lock Release (CRITICAL)

Every code path that acquires a lock MUST release it.

// ALWAYS use try/finally to ensure lock release
function withLock(trackingPath, operation) {
  const tracking = acquireLock(trackingPath);
  try {
    operation(tracking);
  } finally {
    releaseLock(tracking, trackingPath);  // ALWAYS executes
  }
}

Lock Release Points in This Skill:

• After APPROVE decision (Step 3)
• After DENY decision (Step 4)
• On any error that aborts execution

Cumulative Feedback Writing

On DENY, feedback is written to TWO locations:

1. Tracking summary: Appended to tracking.ideas[slug].denial_feedback[]
2. Detailed file: Prepended to [project]/.planning/feedback.md

Both must be updated atomically to maintain consistency.

STATE MACHINE INTEGRATION

Status Transitions

On APPROVE:

On DENY:

Deny Loop Flow

scaffolded_available
        ↓
   /plan-app (MUST read feedback.md)
        ↓
planned_available
        ↓
   /implement-app
        ↓
implemented_available
        ↓
   /verify-app
        ↓
pending_review
        ↓
   /review-app → DENY → scaffolded_available (loop)
              → APPROVE → built (exit)

ERROR HANDLING

Empty Review Queue

REVIEW QUEUE
============

No items pending review.

Status counts:
- scaffolded_available: 3 (need /plan-app)
- planned_available: 2 (need /implement-app)
- implemented_available: 1 (need /verify-app)
- pending_review: 0

Nothing to review. Run the appropriate skill to advance items.

Invalid Slug

ERROR: Idea not found or not pending review

Provided: [slug]
Status: [actual status or "not found"]

Only ideas with status 'pending_review' can be reviewed.
Current review queue:
1. other-idea-1
2. other-idea-2

EXAMPLE: AUTO-DENY FLOW (Default Behavior)

User: /review-million-dollar-app simple-todo-app

Agent:
Starting /review-million-dollar-app for: simple-todo-app
----------------------------------------------------

AUTO-DENY MODE ACTIVE
=====================
Review iteration: #2 of 5 (auto-deny threshold)
Mode: Automatic verification and denial

Spawning verification agent...

[Auto-Deny Agent] Starting comprehensive verification...
[Auto-Deny Agent] Checking infrastructure...
[Auto-Deny Agent] Running build...
[Auto-Deny Agent] Starting dev server...
[Auto-Deny Agent] ERROR: PrismaClient constructor error detected
[Auto-Deny Agent] Running tests...
[Auto-Deny Agent] Tests: 250/250 passed (but app fails to start)

----------------------------------------------------

AUTO-DENY VERIFICATION RESULTS
==============================

| Category          | Status  | Details                                    |
|-------------------|---------|-------------------------------------------|
| Infrastructure    | PASS    | docker-compose.yml valid                  |
| Build             | FAIL    | Prisma client engine error                |
| Runtime           | FAIL    | App returns 404 due to Prisma init error  |
| Tests             | PASS    | 250/250 (mocked DB hides runtime issue)   |
| Code Quality      | WARN    | Engine type 'client' misconfigured        |

Critical Issues Found:
1. [CRITICAL] PrismaClient uses engine type "client" without required adapter
   File: src/lib/db.ts:25
   Fix: Remove 'client' engine or provide adapter/accelerateUrl

2. [CRITICAL] App fails to start - 404 on all routes
   Cause: Prisma initialization fails at module load time

----------------------------------------------------

AUTO-DENIED (Review #2)
=======================

simple-todo-app has been automatically denied.

Tracking Updates:
- Status: scaffolded_available
- Review count: 2
- Feedback: Written to .planning/feedback.md

Auto-Deny Status:
- Iterations remaining: 3 more auto-denies before human review required
- Next review (#3) will also be auto-denied unless --no-auto-deny is used

Next Steps:
1. /plan-million-dollar-app will read feedback and create fix plan
2. /implement-million-dollar-app will apply fixes
3. /verify-million-dollar-app will re-verify
4. /review-million-dollar-app will auto-deny again (or human review at #6)

EXAMPLE: HUMAN REVIEW (After Auto-Deny Threshold)

User: /review-million-dollar-app simple-todo-app

Agent:
Starting /review-million-dollar-app for: simple-todo-app
----------------------------------------------------

HUMAN REVIEW REQUIRED
=====================
Review iteration: #6 (exceeds auto-deny threshold of 5)
Mode: Human decision required

REVIEWING: simple-todo-app
==========================

## Test Results
All 250 tests passing

## App Status
- Build: Success
- Runtime: App starts successfully
- Demo URL: http://localhost:3000

## Prior Denials (5)
Review #1: AUTH_SECRET placeholder, no database setup
Review #2: PrismaClient engine type error
Review #3: Missing environment validation
Review #4: Database migrations not auto-running
Review #5: Auth callback URL misconfigured

All prior issues have been addressed in this iteration.

----------------------------------------------------

DECISION REQUIRED (Human Only)

[A] APPROVE - Mark as 'built' ($1M ready!)
[D] DENY - Return to planning with feedback

Enter decision (A/D):

EXAMPLE: BYPASSING AUTO-DENY

User: /review-million-dollar-app simple-todo-app --no-auto-deny

Agent:
Starting /review-million-dollar-app for: simple-todo-app
----------------------------------------------------

HUMAN REVIEW MODE (--no-auto-deny)
==================================
Review iteration: #2
Note: Auto-deny bypassed by flag

REVIEWING: simple-todo-app
==========================

[... presents full evidence for human decision ...]

DECISION REQUIRED

[A] APPROVE - Mark as 'built' ($1M ready!)
[D] DENY - Return to planning with feedback

Enter decision (A/D):

EXAMPLE: FULL APPROVE FLOW (Human Review)

User: /review-million-dollar-app ai-code-review-tools

Agent:
REVIEWING: ai-code-review-tools
===============================
(Review #6 - human review required)

## Test Results
All 65 tests passing

## Screenshots
5 screenshots in evidence/

## Market Validation
3 sources verified, all current

## Implementation
4/4 phases, 23/23 tasks complete

DECISION REQUIRED
[A] APPROVE [D] DENY

User: A

Agent:
APPROVED!

ai-code-review-tools is now 'built'.
$1M ready!

EXAMPLE: FULL DENY FLOW (Human Review)

User: /review-million-dollar-app smart-inventory-tracker

Agent:
REVIEWING: smart-inventory-tracker
==================================

## Test Results
Unit: 30/32 (2 failed)
Integration: 8/10 (2 failed)

## Screenshots
3 captured (missing settings page)

## PRIOR FEEDBACK (Review #1)
Previously denied for:
- Missing inventory import feature
- No CSV export

DECISION REQUIRED
[A] APPROVE [D] DENY

User: D

Agent:
DENIAL FEEDBACK
===============

## Issues Found
> 1. Tests still failing
> 2. Import feature incomplete
> 3. Export only works for small datasets

## Required Changes
> 1. Fix failing tests
> 2. Complete CSV import feature
> 3. Add pagination to export

Confirm DENY? (Y/N)

User: Y

Agent:
DENIED

smart-inventory-tracker returned to scaffolded_available.
Review count: 3
Feedback appended to .planning/feedback.md

/plan-app MUST address all 3 reviews' feedback.

VERIFICATION

After this skill completes, verify:

On APPROVE

1. Tracking status updated to built
2. Item removed from review_queue
3. History entry added for approval
4. approved_at timestamp set

On DENY

1. Tracking status updated to scaffolded_available
2. review_count incremented
3. Feedback summary appended to denial_feedback[] array
4. Item removed from review_queue
5. History entry added for denial
6. .planning/feedback.md exists in project directory
7. New feedback prepended to .planning/feedback.md
8. All prior feedback preserved in .planning/feedback.md

NEXT STEPS AFTER COMPLETION

After APPROVE

• Celebrate! The app is $1M ready
• Status is built (terminal success)
• No further action required

After DENY

1. Run /plan-million-dollar-app [slug]
   • Agent MUST read .planning/feedback.md
   • New plan MUST address ALL prior feedback
2. Run /implement-million-dollar-app [slug]
3. Run /verify-million-dollar-app [slug]
4. Return to /review-million-dollar-app

The cycle continues until APPROVED or manually ABANDONED.