ads-creative-analysis
Skill1 starsChanged 23 days agoMITView on GitHub
Meta-first creative intelligence - hooks, first seconds, messaging, offer, CTA, visual format, UGC vs polished, static vs video, fatigue, audience-message fit, pattern interrupts, video retention - and structured creative briefs plus single-variable creative tests. Use for "which creative is winning", "new ad ideas", "creative fatigue", "write ad copy", "brief for my video team".
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- ads-auditSkill
- ads-budgetingSkill
- ads-client-onboardingSkill
- ads-communitySkill
- ads-competitor-researchSkill
- ads-coreSkill
- ads-landing-pageSkill
- ads-optimizationSkill
- ads-portfolioSkill
- ads-reportingSkill
- ads-strategySkill
- ads-trackingSkill
- google-adsSkill
- meta-adsSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 29
Do not edit or render video.
- Distinctive: line 43
Use the client's brand voice and service area.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Creative analysis
Analyze
For each ad with meaningful spend (≥ ~1x target CPA, HEURISTIC), within its ad set:
| Dimension | Read from |
| Hook / first 3 s | hook rate (3 s plays / impressions); opening frame; first line of primary text |
| Retention | hold rate, p25-p100 curve - where people leave |
| Message | the promise, pain, proof, objection handled |
| Offer + CTA | link CTR, CTA button, offer clarity |
| Format | static / video / carousel; UGC / founder / polished; aspect ratios available |
| Fatigue | frequency, CTR trend, age of ad (fatigue signature) |
| Audience-message fit | does the creative speak to the target customer in the profile? |
| Downstream | CVR and lead quality (CRM) - a high-CTR ad can drive bad leads |
Compare ads within the same ad set and window; Meta concentrates spend on predicted winners, so low-spend ads are "untested", not "losers". Use adsgm evidence grading before calling a winner.
Tag each ad with a short taxonomy (hook type: question / bold claim / before-after / problem / demo / testimonial; format; talent; offer) so patterns across ads become visible.
Produce briefs, not videos
Do not edit or render video. Output structured briefs the user's content/video workflow consumes: templates/creative-brief.md. Every brief has: hook, script/message, desired length, format, B-roll, CTA, caption, visual references, platform/placement, hypothesis. 3-5 hook variants per body are cheap and high-leverage.
Graphic ads
Turn briefs into on-brand static images with adsgm creative render --brief brief.json:
- templates
offer-card,photo-overlay,quote,split; - formats 1x1, 4x5, 9x16 and 1.91x1;
- uses the profile's
brandcolors, font and logo; - blocks prohibited claims.
It writes SVG, and PNG when Edge or Chrome is available, plus a contact sheet. See workflows/graphic-ads.md. Only real, client-supplied photos and reviews. Uploads go in PAUSED.
Copy rules
- Use the client's brand voice and service area; no prohibited claims; no invented proof (reviews, awards, years in business, prices, financing).
- Meta: primary text hook within the first ~125 characters; headline short; one CTA.
- Google RSA: headlines ≤30 chars, descriptions ≤90; include keyword theme, location, offer, trust, CTA.
Testing
workflows/creative-testing.md - one variable per test, logged in templates/experiment-log.json, read at ≥2x target CPA per variant.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.