| name | ad-script-generator |
| description | Generate 100+ ad script variations from competitor intelligence. Takes ad-research output, rewrites competitor angles in brand voice, multiplies across ICPs, and produces production-ready scripts for UGC creators and AI video (Arcads). Use when creating ad scripts, generating creative variations, writing ad copy at scale, or building a creative testing pipeline. Mirrors Eddie's Stage 3 (Script Generation). |
| risk | low |
| tags | ["growth","marketing","content","creative"] |
| created | 2026-03-23T00:00:00.000Z |
| updated | 2026-03-23T00:00:00.000Z |
Ad Script Generator — Volume Creative Engine
Generate dozens to hundreds of ad script variations from proven competitor patterns, rewritten in your brand voice and multiplied across ICPs. This is UAOP Stage 3 (Variation Generation) for the Growth department.
When to use
- After running
ad-research and having a competitive intel dump
- When you need fresh ad creatives for testing
- When existing ads are burning out (fatigue after 4-6 weeks)
- When expanding to a new ICP or product surface
- Monthly creative refresh cycles
When NOT to use
- Competitor research (use
ad-research first)
- Performance analysis of running ads (use
ad-performance-loop)
- Brand strategy or positioning work (use
brand-guidelines or marketing-strategy-pmm)
Prerequisites
Before running, ensure these files are loaded:
REQUIRED CONTEXT:
1. ~/.claude/contexts/writing-rules.md (anti-slop filter)
2. ~/.claude/contexts/growth/voice.md (brand communication style)
3. ~/.claude/contexts/growth/rules.md (growth invariants)
4. ~/.claude/contexts/growth/products/{surface}.md (product facts)
5. ~/.claude/contexts/growth/icps/{icp}.md (ideal customer profiles)
6. Ad research intel dump (from ad-research skill)
Pipeline
Step 1: Ingest Competitor Intel
Load the competitive ad intel dump from ad-research. For each competitor ad:
- Extract the hook, angle, structure, and body copy
- Rate reusability: High (proven angle, easy to adapt) / Medium / Low
- Filter to High + Medium reusability ads only
Step 2: Rewrite in Brand Voice
For EACH high-reusability competitor ad, produce 2 outputs:
Output A — Original Analysis:
## Original: [Competitor] Ad #[N]
**Hook:** "[exact hook]"
**Angle:** [angle type]
**Structure:** [hook → problem → solution → proof → CTA]
**Why it works:** [specific analysis]
**Body copy:** "[exact copy]"
**Transcript:** "[if video]"
Output B — Brand Rewrite:
## Rewrite: [Our Product] — Based on [Competitor] Ad #[N]
**Hook:** "[rewritten hook in our voice]"
**Angle:** [same angle, our positioning]
**Structure:** [same structure, our product facts]
### Script (Video — 30 seconds)
[HOOK — 0-3s]
"[opening line — must stop the scroll]"
[PROBLEM — 3-8s]
"[pain point in ICP's language]"
[SOLUTION — 8-18s]
"[our product solving it — specific features/benefits]"
[PROOF — 18-25s]
"[testimonial/stat/demo]"
[CTA — 25-30s]
"[clear call to action]"
### Body Copy (for static/carousel)
Headline: "[headline]"
Primary text: "[body copy]"
CTA: "[button text]"
Step 3: ICP Multiplier
Take each rewrite and multiply across ICPs:
For each rewrite:
For each ICP in ~/.claude/contexts/growth/icps/:
→ Adapt language to ICP's vocabulary
→ Swap pain points to ICP-specific problems
→ Adjust proof points to what ICP cares about
→ Keep the same angle and structure
Example: If you have 20 rewrites and 3 ICPs, you get 60 scripts.
Step 4: Variation Generation
For top-performing angles (identified by ad-performance-loop or human judgment):
For each top script:
→ Generate 3-5 hook variations (different opening lines, same angle)
→ Generate 2-3 length variations (15s, 30s, 60s)
→ Generate 2 tone variations (serious vs. casual)
This multiplies 60 scripts into 100-300+ variations.
Step 5: Production Routing (UAOP Stage 4)
Sort scripts into two pipelines:
Pipeline A — UGC Creators (Top 10-15):
- Highest-conviction scripts (proven angle + strong hook + clear ICP fit)
- Include: script + reference competitor video + brand voice guide
- Budget: $15-50 per video
- Format: Creator brief with shot list
Pipeline B — AI Video (Remaining):
- All other scripts go to Arcads (or equivalent) via API
- Same script → 3-5 different AI actors
- Same actor → 5-10 different hooks
- Output: 50+ finished video ads per batch
Output Format
Script Document
# Ad Script Batch — [Date]
## Batch Summary
- Competitor ads analyzed: [N]
- Rewrites generated: [N]
- ICP multiplication: [N] ICPs × [N] rewrites = [N] scripts
- Hook variations: [N]
- Total scripts: [N]
## Pipeline A — UGC Creator Scripts (Top [N])
[numbered list with full scripts + reference videos]
## Pipeline B — AI Video Scripts (Remaining [N])
[numbered list with full scripts + actor recommendations]
## Angle Breakdown
| Angle | Scripts Using | Source Competitors | Expected Performance |
|-------|--------------|-------------------|---------------------|
Anti-Slop Rules (Mandatory)
Before finalizing ANY script:
- Run through
anti-ai-writing banned word list — zero tolerance
- Check against
growth/voice.md — must sound human
- Verify product facts against
growth/products/{surface}.md
- Ensure ICP language matches
growth/icps/{icp}.md
- Read the script aloud — if it sounds like an AI wrote it, rewrite it
Tools Used
| Tool | Purpose |
|---|
| memory MCP | Load competitor intel, store winning scripts |
| sequential MCP | Multi-step reasoning for rewrites |
| filesystem | Read context files (voice, product, ICP) |
Agents
| Agent | Role |
|---|
| growth-marketer (L2) | Owns script strategy, selects top scripts |
| content-strategist (L3) | Writes rewrites, ensures voice consistency |
| prompt-engineer (L3) | Optimizes script templates for quality |
| analytics-reporter (L3) | Tracks which angles produce winners |