content-review

name	content-review
description	Review content plans, scripts, image prompts, video prompts, and thumbnails BEFORE generating to catch issues early and save money. Use before any expensive generation. Keywords - review prompt, check script, validate plan, review thumbnail, review before generating, save money, catch errors.
allowed-tools	Bash Read Write Edit Glob Grep

Content Review (Pre-Generation)

Review content BEFORE spending money on generation. Catch issues in plans, scripts, and prompts early - a 1-second review can save $0.07-$20+ in wasted generations.

All review functions are available via the CLI:

node workflows/cli.cjs list   # show all commands (reviewScript, reviewImagePrompt, reviewVideoPrompt, reviewThumbnail, reviewContentPlan, reviewBatch, ...)

---

Why Review First?

Content Type	Generation Cost	Review Cost
Script/Hook	$0.05-0.50 (TTS)	FREE
Thumbnail	$0.067-0.15	FREE
Image	$0.067-0.15	FREE
Video (30s)	$0.90-6.00	FREE
Carousel (10 slides)	$0.67	FREE

One failed video = 10-100x the cost of reviewing first.

---

What Gets Caught

Script Review (reviewScript)

Issue	Severity	Why It Matters
Restricted words (guarantee, cure)	ERROR	Brand/legal violations
Script too long for duration	WARNING	Rushed delivery, bad pacing
Script too short	WARNING	Awkward pauses, dead air
No clear CTA	SUGGESTION	Missed conversion opportunity
Long hook (first sentence)	SUGGESTION	Loses attention in first 3s

Image Prompt Review (reviewImagePrompt)

Issue	Severity	Why It Matters
Text in prompt	ERROR	AI generates garbled text
Logo in prompt	ERROR	Logos render poorly
Animation terms	ERROR	Images are static
No style keywords	SUGGESTION	Generic results
No lighting direction	SUGGESTION	Flat lighting
Person without reference	WARNING	Inconsistent character

Video Prompt Review (reviewVideoPrompt)

Issue	Severity	Why It Matters
Text/captions in prompt	ERROR	Garbled on-screen text
Conflicting camera moves	ERROR	Impossible to render
Multiple scenes in one prompt	ERROR	Use multi-shot workflow
Transitions in prompt	ERROR	Belongs in editing
No camera movement	SUGGESTION	Static, boring footage
Veo + speaking character	WARNING	No native lip-sync
Seedance + no dialogue	ERROR	Lip-sync needs dialogue

Thumbnail Review (reviewThumbnail) - NEW

Issue	Severity	Why It Matters
Person but no character reference	ERROR	Inconsistent face across content
Text/logo in prompt	ERROR	AI generates garbled text
Small face	WARNING	Unrecognizable at thumbnail size
Multiple faces	WARNING	Confusing at small size
Busy background	WARNING	Reduces thumbnail impact
No expression/emotion	WARNING	Boring, no curiosity
Video thumbnail mismatch	ERROR	Must match video character
No high-contrast colors	SUGGESTION	Doesn't stand out in feed
No text space	SUGGESTION	No room for post-production text

Special Output:

• recommendedWorkflow: 'generateImageVariation' or 'generateSingleImage'
• characterStatus: 'has-reference', 'needs-reference', or 'not-needed'
• nextSteps: Array of actions to take

Content Plan Review (reviewContentPlan)

Issue	Severity	Why It Matters
Unfilled placeholders	ERROR	Incomplete plan
Missing required sections	WARNING	Incomplete brief
Brand restriction violation	ERROR	Off-brand content
No cost estimate	SUGGESTION	Budget tracking

---

How to Use

1. Review a Script Before TTS/Video

node workflows/cli.cjs reviewScript '{"script":"This product will cure your problems guaranteed!","targetDuration":15,"platform":"tiktok","restrictions":["cure","guaranteed","risk-free"],"contentType":"ad"}'
# Returns { data: { pass, issues: [{ severity, message }], potentialSavings } }
# If data.pass is false, STOP and fix issues before running generateVoiceover.

2. Review an Image Prompt Before Generation

node workflows/cli.cjs reviewImagePrompt '{"prompt":"Professional woman holding product with price tag showing $29","aspectRatio":"9:16","restrictions":["price","discount"],"expectsPerson":true,"characterDescription":"Woman, 30s, dark hair, business casual"}'
# Returns { data: { pass, summary } }; if data.pass is false, fix issues before generating.

3. Review a Video Prompt Before Generation

node workflows/cli.cjs reviewVideoPrompt '{"prompt":"Camera zooms in while panning left, text appears showing \"50% OFF\"","duration":8,"provider":"veo","restrictions":["text","discount"]}'
# data.pass === false; issues:
#   - [ERROR] Text in prompts will be garbled
#   - [ERROR] Conflicting camera movements
#   - [ERROR] Contains restricted word: "discount"

4. Review a Thumbnail Before Generation (IMPORTANT)

# hasCharacterReference false = no reference provided; forVideo true = this is for a video
node workflows/cli.cjs reviewThumbnail '{"prompt":"Person with excited expression on blue background","thumbnailType":"youtube","includesPerson":true,"hasCharacterReference":false,"forVideo":true}'
# data.pass === false
# data.characterStatus === 'needs-reference'
# data.nextSteps === [
#   '1. Generate character keyframe: generateSingleImage() with detailed description',
#   '2. Register in asset registry: registerAsset()',
#   '3. Generate thumbnail: generateImageVariation() with reference'
# ]

Thumbnail with character reference (correct):

node workflows/cli.cjs reviewThumbnail '{"prompt":"Same person with shocked expression, mouth open, high contrast","thumbnailType":"youtube","includesPerson":true,"hasCharacterReference":true,"characterReferencePath":"projects/my-project/assets/characters/char-main.png","forVideo":true,"videoCharacterPath":"projects/my-project/assets/characters/char-main.png"}'
# data.pass === true
# data.recommendedWorkflow === 'generateImageVariation'
# data.nextSteps === ['Use generateImageVariation() with referenceImagePath']

4. Review a Content Plan Before Generation

node workflows/cli.cjs reviewContentPlan '{"planPath":"projects/my-project/content-plans/day-01/content-01-video.md","brandPath":"projects/my-project/templates/brand.md","contentType":"video"}'
# Returns { data: { pass, issues } }; if data.pass is false, the plan needs fixes.

5. Batch Review Multiple Prompts

node workflows/cli.cjs reviewBatch '{"imagePrompts":["Product hero shot, studio lighting","Woman using product with text showing features"],"videoPrompts":["Camera dollies in on product, premium feel","Character speaks to camera with caption overlay"],"scripts":["This will cure your problems...","Try it free, link in bio."],"restrictions":["cure","guaranteed","caption"]}'
# Returns { data: { pass, averageScore, totalPotentialSavings } }

---

Workflow Integration

Before Image Generation

1. Write image prompt
2. reviewImagePrompt() ← REVIEW
3. Fix any errors
4. generateSingleImage() or generateImageVariation()

Before Video Generation

1. Write video prompt
2. reviewVideoPrompt() ← REVIEW
3. Fix any errors
4. generateSingleImage() keyframe (preview) — or generateStoryboard() for multi-scene
5. reviewOutput() (QA the keyframe)
6. generateVideoFromImage() / generateSpeakingVideoFromImage()

Before Thumbnail Generation (CRITICAL)

1. ASK: "Does thumbnail include a person?"
2. If YES: Check asset registry for character reference
3. reviewThumbnail() ← REVIEW (checks reference requirement)
4. If needs-reference:
   a. Generate character keyframe first
   b. Register in asset registry
   c. THEN generate thumbnail with reference
5. If has-reference: generateImageVariation()
6. If faceless: generateSingleImage()

Before TTS/Voiceover

1. Write script
2. reviewScript() ← REVIEW
3. Fix any errors (pacing, restrictions)
4. generateVoiceover()

Before Campaign Generation

1. Create content plan
2. reviewContentPlan() ← REVIEW
3. Fix any errors
4. Extract prompts/scripts
5. reviewBatch() ← REVIEW ALL
6. Generate content

---

Review Result Format

All review functions return:

{
  pass: boolean,           // false if any errors
  score: number,           // 0-100
  issues: [{
    category: string,      // 'restrictions', 'prompt-quality', 'pacing', etc.
    severity: 'error' | 'warning' | 'suggestion',
    message: string,       // What's wrong
    suggestion?: string,   // How to fix
    location?: string,     // Where in the content
  }],
  summary: string,         // Human-readable summary
  potentialSavings: number // $ saved by fixing before generating
}

Thumbnail review returns additional fields:

{
  ...baseFields,
  recommendedWorkflow: 'generateImageVariation' | 'generateSingleImage',
  characterStatus: 'has-reference' | 'needs-reference' | 'not-needed',
  nextSteps: string[]      // Step-by-step actions to take
}

---

Acting on Results

pass === true   → Safe to generate
pass === false  → STOP. Fix errors first.

For each issue:
  severity === 'error'      → MUST fix before generating
  severity === 'warning'    → Should fix, user can override
  severity === 'suggestion' → Nice to have, optional

Always show the user:

1. The issues list
2. The potential savings
3. Suggested fixes

---

Cost Savings Examples

Scenario	Without Review	With Review
Text in 5 image prompts	$0.34 wasted	$0 (caught)
Conflicting camera in 30s video	$2.40 wasted	$0 (caught)
Restriction violation in script	$0.05 + $6 video	$0 (caught)
Thumbnail without character ref	$0.10 + inconsistent branding	$0 (caught)
Incomplete plan → bad outputs	$10+ wasted	$0 (caught)

Review costs pennies (text-only LLM calls, budget-gated). Generation costs real money.

---

Quick Reference

Function	Input	Checks
reviewScript	script text, duration, restrictions	Pacing, restrictions, hook, CTA
reviewImagePrompt	prompt, aspect, restrictions	Text, animation, style, lighting
reviewVideoPrompt	prompt, duration, provider	Text, camera, transitions, provider fit
reviewThumbnail	prompt, type, includesPerson, hasRef	Character ref, text/logo, expression, contrast
reviewContentPlan	plan path, brand path	Completeness, placeholders, brand
reviewBatch	arrays of prompts/scripts	All of the above