/validate-production

Parse Arguments

From $ARGUMENTS, extract:

• --input (required): Path to video prompt JSON file or script JSON file to validate
• --checks (optional, default: all): Which validation checks to run (all, constraints, continuity, complexity, cost)
• --format (optional, default: both): Output format (json, md, both)

If --input is missing, ask for the file path.

Step 1: Load Skill

Read ${CLAUDE_PLUGIN_ROOT}/skills/video-validator/SKILL.md

Step 2: Load Input File

Read the input file at the provided path. Determine file type:

• If contains ## PROMPT METADATA or segments with prompt_text: it's a prompt output
• If contains ## SCRIPT METADATA or segments with full_script: it's a script output
• If JSON: parse and determine from schema

Step 3: Run Validation Checks

If --checks includes constraints or all:

Run platform constraint validation:

python3 ${CLAUDE_PLUGIN_ROOT}/lib/platform_constraints.py

Check each segment:

• Duration within platform limits
• Aspect ratio is valid for platform
• Resolution is valid for platform
• Distribution platform limits (if specified)

If --checks includes continuity or all:

Run continuity checker (only for multi-segment outputs):

python3 ${CLAUDE_PLUGIN_ROOT}/lib/continuity_checker.py

Compare character sheet fields across all segments:

• CRITICAL: wardrobe, hair, facial features, age
• WARNING: lighting, color temperature, background style
• INFO: expression, gesture style, energy level

If --checks includes complexity or all:

Score each prompt for generation complexity:

• Number of distinct subjects (each +0.1)
• Number of simultaneous actions (each +0.15)
• Camera movement complexity (0.0-0.2)
• Lighting specificity (0.0-0.1)
• Text/logo presence (+0.3)

Flag:

• Score 0.5-0.7: WARNING — may need simplification
• Score >0.7: FAIL — recommend splitting or simplifying

Check content policy:

• No text/logos/brands in visual descriptions
• No real person names without character system
• No adult or restricted content themes

If --checks includes cost or all:

Run cost estimation:

python3 ${CLAUDE_PLUGIN_ROOT}/lib/cost_estimator.py --segments '[segment data]' --format json

Calculate:

• Individual generation cost (per segment)
• Batch generation cost (grouped segments)
• Savings percentage
• Optimal strategy recommendation

Step 4: Generate Validation Report

Output using structured format:

## VALIDATION REPORT

**Input File:** [path]
**File Type:** [Script | Prompt]
**Segments:** [N]
**Platform:** [Sora 2 | Veo 3.1 | Both]
**Checks Run:** [list of checks]

---

## PLATFORM CONSTRAINTS

### Segment 1
- Duration: [PASS/FAIL] — [Xs] (valid: [list])
- Aspect Ratio: [PASS/FAIL] — [ratio] (valid: [list])
- Resolution: [PASS/FAIL] — [res] (valid: [list])

[Additional segments...]

### Distribution Limits
- [Platform]: [PASS/FAIL] — [duration]s at [ratio]

---

## CHARACTER CONTINUITY

### Cross-Segment Comparison
[PASS if single segment or all fields match]

| Field | Seg 1 | Seg 2 | Seg 3 | Seg 4 | Status |
|-------|-------|-------|-------|-------|--------|
[Field comparison table]

### Issues Found
- [CRITICAL/WARNING/INFO]: [description]

---

## PROMPT COMPLEXITY

### Segment 1
- Score: [X.XX] / 1.0
- Rating: [Simple/Moderate/Complex/Too Complex]
- Subjects: [N]
- Actions: [N]
- Recommendation: [if needed]

[Additional segments...]

### Content Policy
- [PASS/WARN]: [details]

---

## COST ESTIMATE

### Individual Generation
| Segment | Duration | Platform | Credits |
|---------|----------|----------|---------|
[table]

**Total:** [X] credits / [Y] API calls

### Batch Generation
**Total:** [X] credits / [Y] calls
**Savings:** [Z]%

### Recommendation
[Strategy recommendation]

---

## SUMMARY

**Overall:** [PASS / FAIL]
**Errors:** [N]
**Warnings:** [N]
**Info:** [N]

### Action Items
1. [Required fix]
2. [Recommended improvement]

Step 5: Output

If --format is md or both:

Write report to video_validation_report.md

If --format is json or both:

Write structured JSON to video_validation_report.json

If --format is both:

Write both files and display summary.