Generate high-performing ad creative at scale
ad-creativeskillsetup L2★37,571
sickn33/antigravity-awesome-skills ↗Causal-lift measurements
ad-copy-generation-8pp vs no-skill baselinewith-skill 88% · baseline 96%
Measured by running the task with and without this artifact, K=5, graded by deterministic checks — no LLM judging.
What it does
Generate high-performing ad creative headlines and descriptions at scale
Best for
Paid media teams managing Google Ads, Meta, or LinkedIn campaigns who need to test new creative variants informed by what's currently working, not guessing.
Inputs
- · Platform and format (Google RSA, Meta feed, LinkedIn, TikTok, Twitter/X)
- · Product/offer and value proposition
- · Target audience and decision stage (problem-aware, solution-aware, product-aware)
- · Performance data from running ads (optional; if iterating)
Outputs
- · Headlines (platform-specific character limits)
- · Descriptions or primary text
- · Platform-compliant ad copy set ready to test
Requires
- · Product marketing context (optional, from .claude/product-marketing-context.md)
- · Performance analytics (CSV, API, or pasted)
Preconditions
- · Platform and format specified
- · Product positioning known
Failure modes
- · Exceeds platform character limits (copy rejected)
- · Iterates without analyzing performance data (shoots in dark)
- · Generic adjectives ('cutting-edge', 'revolutionary') without proof
- · No hypothesis on what's winning (data-free iteration)
Trust signals
- · Platform specs table with exact character limits per element (Google: 30 char headlines, 90 char descriptions; Meta: 125 chars primary, 40 char headline)
- · RSA rules documented (headlines must work independently, pin to position only when necessary)
- · Iteration loop: pull data → identify patterns → generate variants → validate specs → deliver
- · Performance data integration (CSV, API, or pasted)