matteotitta Ad Creative Brief
Generate creative briefs for ad designers and video teams. Extracts messaging angles from copy skill outputs, adds visual direction, format specs per platform, and A/B test variants. Outputs a structured brief ready for design production.
Doctrine inherited (Step 7 — 0626 rollout, locked 2026-06-04)
Output complies with:
output-tenets.md,output-simplicity.md,design-production.md- Step 6 calibration: see [[feedback_execution_doctrine_refinements_step6]]
Refinements applied to this skill:
| Code | Refinement | How it lands in ad-creative-brief |
|---|---|---|
| R1 | Source placement | Brief is client-team review surface (designer hand-off). Cleaned [VERIFIED:...] tags in a collapsed appendix. No tags on the per-platform spec sheets. |
| R2 | Single-doc-with-toggles | Multi-platform brief ships as one doc with toggle per platform (LinkedIn 1.91:1, LinkedIn 1:1, Meta 1:1, Meta 4:5, Meta 9:16) — not 5 sub-docs. |
| R3 | Product-update tone | Ad copy directions frame as "[Product] does X" not "thrilled to introduce." |
| R5 | Blog as voice anchor | When the campaign has an anchor blog, the brief specifies the blog's opening line as the hero headline across all formats. |
| R6 | CTA hierarchy | Market-facing ads → sign-up primary, blog as fallback. Retargeting / warm-base → product-action. |
| R9 | Action-oriented section names | "Strategy / Hero overlay / Visual direction" — action/noun-led. |
Triggers
Run this skill when:
- Ad copy is written (via
/google-ads-copyor/linkedin-ads-copy) and creative production needs to start - A designer or video producer is being briefed for new campaign creative
- Existing creative is fatigued and needs refresh angles before re-shoot
- A new platform/format is being added to an existing campaign
Do NOT run when:
- Copy isn't written — run
/google-ads-copyor/linkedin-ads-copyfirst - The ask is a single image (not a brief) — use a designer directly
- DESIGN.md doesn't exist — pause and run
/brand-kit
The Iron Law — voice-locked
VISUAL DIRECTION IS NOT A MOCKUP.
The creative brief bridges copy and design. It tells the designer what the ad should communicate and what it should feel like — not pixel-perfect layouts. Overly prescriptive briefs kill creative quality. Under-specified briefs waste revision cycles.
No exceptions:
- "Make it blue with the logo top-left" → That's art direction, not a brief. Describe the feeling and message, let the designer compose.
- "Just use the copy as-is on a background" → Every format needs a visual concept. Text-on-gradient is not a concept.
- "We need 20 variants" → Test fewer, learn faster. 2-3 variants per test variable. Start with concept/angle, not color swaps.
- "Skip the specs, the designer knows" → Platform specs change constantly. Always include dimensions + file size limits. Google will reject non-compliant assets.
DESIGN.md token contract — voice-locked
Every brief consumes the client's DESIGN.md at projects/consulting/{client}/brand/{MMYY}-brand-kit.md and cites tokens explicitly. Same brand contract as the rest of the marketing surface.
| Required in every brief | Forbidden |
|---|---|
| DESIGN.md path referenced at top | Color descriptions without token references |
Token-cited colors (colors.primary etc.) |
Type styles outside typography.* tokens |
Token-cited type (typography.headline-lg etc.) |
Off-brand corner radii, custom font sizes |
Component specs cited (components.button-primary) |
Ad treatments that compete visually with brand surfaces |
| Do's/Don'ts as guardrails (one primary per screen, two weights max) | Inventing colors or fonts when DESIGN.md doesn't have them |
Authority: Full integration contract in .claude/rules/design-production.md (auto-loaded). When in doubt, that file wins.
If no DESIGN.md exists: pause and recommend running /brand-kit first. Do not invent tokens.
A/B test hierarchy — voice-locked
Test in order. Each level has diminishing returns compared to the one above:
| Priority | Variable | What to test |
|---|---|---|
| 1 (mandatory first) | Concept/angle | Pain point vs outcome vs social proof |
| 2 | Layout/composition | Image left vs right, text overlay vs clean |
| 3 | CTA | Direct vs soft vs proof-led |
| 4 (lowest) | Color/style | Photo vs illustration |
Never test color before concept. Concept gets the largest delta in performance; color almost never moves the needle when the angle is wrong.
Which concept to test first is the angle matrix's job ([the premium reference](the premium reference)) — it scores the candidate angles by evidence so the mandatory first test starts from the highest-opportunity cell, not a guess.
Inputs
| Input | Description | Required |
|---|---|---|
| google-ads-copy or linkedin-ads-copy | Source copy with headlines, descriptions, formats | Required (at least one) |
| product-messaging | Value props, differentiators | Required |
| paid-campaign-strategy | Campaign architecture | Required |
| DESIGN.md (brand-kit) | Tokens for color, type, components | Required |
| tov-guidelines | Voice patterns for hero text overlays | Recommended |
| linkedin-ad-teardown | Competitor gaps → the matrix's competition-density axis | Recommended |
Process
Six-phase flow: Platform/format identification → the angle matrix (score angle × persona × awareness on evidence, brief the top band) → Visual direction per angle → Format specs per platform → A/B test plan → Structured brief output. Full flowchart, image specs per platform (Google Display + LinkedIn), test rules, and brief quality standards in the premium reference; the scored-matrix method in [the premium reference](the premium reference).
Anti-Hallucination Guardrails
- Never invent customer stats. "2,000 advisers trust us" needs a verified source. Otherwise mark
[NOT AVAILABLE]and remove. - Never invent token values. If DESIGN.md doesn't define a needed color, pause — don't guess.
- Never approximate platform specs. Use the table in
process.mdor pull current values from platform docs. - Never collapse angles. If you can only generate 2 distinct angles, say so — don't pad with paraphrases.
- Never strip the caption requirement from video briefs. 85% of LinkedIn video is watched muted; captions are mandatory.
Quality
Self-evaluation checklist (token coverage, distinct angles, spec accuracy, test ordering), worked example (suitability-report time-savings), anti-examples (orphan hex, pixel art direction, paraphrase angles), failure-mode triage in the premium reference.
Integration with Other Skills
Upstream (consumes)
| Skill | What it provides | Required? |
|---|---|---|
google-ads-copy or linkedin-ads-copy |
Source copy | Required (at least one) |
product-messaging |
Value props, proof points | Required |
paid-campaign-strategy |
Campaign architecture | Required |
brand-kit |
DESIGN.md tokens | Required |
tov-guidelines |
Voice patterns for overlays | Recommended |
Downstream
Terminal deliverable — output goes directly to a designer or video producer. No Genesys-internal downstream skill consumes this output.
Final ship gate
Run /premortem --output before ship. See /premortem skill for the 5 execution domains (will-it-resonate / will-it-convert / will-it-stay-on-brand / will-stakeholder-push-back / will-it-degrade-over-time) and output template.
Trivial-case escape: ## Premortem\nNo failure modes — trivial change satisfies the contract for genuinely trivial outputs.
Persuasion & stickiness pass
Output complies with persuasion-and-stickiness.md — Cialdini's 7 persuasion levers + Heath's SUCCESs. Deploy the 1-2 Cialdini levers that fit the reader's barrier (never all seven; every lever must be TRUE), run the SUCCESs diagnostic (Simple / Unexpected / Concrete / Credible / Emotional / Stories) over the near-final draft, then the rule's pre-ship gate.