Creative Production Ads Explorer

Ads Explorer

Use this skill when the user wants to see many different image-ad directions from one subject anchor: product reference image, packaging, service, venue, app, campaign offer, or business asset.

This is the Explore-stage owner for ad-format diversity. It is intentionally broader than scene-explorer and more ad-specific than offer-explorer.

Read references/experience-contract.md before writing the user-facing handoff.

Read references/artifact-contracts.md before creating, repairing, or reporting artifacts.

Read references/codex-exec-image-generation.md before running image generation.

Read references/image-building-strategy.md alongside the Codex exec contract for first-pass imagery rules.

Pipeline Role

Use ads-explorer when the request is about ad formats, campaign images, paid social ideas, launch ads, proof-led ads, ecommerce ads, OOH mockups, UGC thumbnails, surreal product posters, or a visual wall of ad directions.

Use scene-explorer first when the main question is where the product should appear in the real world.

After the review wall, keep more versions in the same mood-board surface. Use tile-level Remix controls, item remixSuggestions, or follow-up generation for controlled changes such as different character, scene, copy, format, crop, style, palette, camera, props, or product placement, then append new results to the saved board.

Default Library

Use the checked-in packs under assets/image-ad-library/packs/.

The default diverse-image-ad-archetypes pack contains 25 reusable starter ad families. The number describes starter directions in the pack, not a required mood-board tile count. It declares:

• per-family prompt structure;
• per-family image size where a non-square crop is better;
• review-wall behavior with captions suppressed;
• copy safety rules for exact headline, CTA, claims, proof, and placeholder text.

Use --pack digital-product-core-ad-prompts for app, SaaS, AI product, marketplace, commerce, developer-tool, or other software-product ads where the interface should be the product proof.

Workflow

1. Capture the subject anchor.

   • If the user supplies an image, preserve exact visible product facts: silhouette, colors, materials, labels, logos, markings, proportions, and use context.
   • If the user supplies a brief only, extract the same facts from the brief.
   • Capture approved copy, headline, CTA, claims, proof, audience, palette, channel, and avoid list.

2. Build the ad prompt wall.

   • Use the shared prompt-pack engine with the ad library:

python3 plugins/creative-production/skills/ads-explorer/scripts/build_ads_explorer.py \
  --ad-name "<subject name>" \
  --subject-kind product \
  --ad-brief "<facts to preserve, approved copy, audience, palette, and avoid list>" \
  --out-dir outputs/imagegen/<subject-slug>-diverse-image-ads

For software-product ads:

python3 plugins/creative-production/skills/ads-explorer/scripts/build_ads_explorer.py \
  --ad-name "<app or digital product name>" \
  --pack digital-product-core-ad-prompts \
  --subject-kind digital-product \
  --ad-brief "<product category, UI surface, job to be shown, approved copy, audience, palette, and privacy constraints>" \
  --out-dir outputs/imagegen/<subject-slug>-digital-product-ads

3. Generate images when requested.

   • Use Codex exec fanout from references/codex-exec-image-generation.md.
   • If a source image is supplied, include its path and preservation requirements in job metadata and prompts.
   • If no source image is available, use text-only generation from the prompt manifest.
   • Do not generate one image for every starter family unless the user asks for generation or clearly wants a full review wall.

4. Render the review wall.

   • The default review should be image-led: a wall of ads, no subtitles or captions under tiles.
   • Prompt details, family names, and sizing metadata should stay in prompts-manifest.json and visual-explorer-metadata.json.
   • Use mixed sizes from the pack. Do not force every ad to square.
   • When an inline MCP review surface is available, render the generated ad set with render_moodboard_board_widget.

5. Handoff.

   • Identify strongest ad families and any format fit issues.
   • Note copy or product-fidelity risks.
   • Use the same lightweight handoff wording as the other explorers: selected direction, preserve, avoid, focused next owner, and artifact path.

Copy Rules

• Use exact supplied text only.
• Avoid subtitles, explanatory body-copy captions, fake price tags, fake reviews, invented proof, and unsupplied CTA text.
• For proof, comparison, certification, ecommerce, and event formats, use placeholder zones unless the brief supplies real details.
• Treat generated text as directional. Final publish assets need deterministic type/layout.
• When the ad brief touches regulated or high-risk categories, add a compact user-facing caveat only when requested, when the category is clearly relevant, or when the asset is moving toward publish use. High-risk categories include prescription drugs or telehealth, supplements or weight loss, financial products, housing, jobs, credit, education, alcohol, tobacco, cannabis, gambling, children-directed products, privacy or security claims, endorsements, and testimonials. Use only approved claims and required disclosures; if disclaimer copy must appear in the asset, ask for the exact approved text.

Exit Criteria

A successful Ads Explorer run ends with:

• up to 25 starter ad prompt directions, with generated ad images only for the families selected or requested;
• an image-led review wall;
• no caption-heavy review framing;
• mixed aspect ratios where useful;
• product facts and copy constraints preserved;
• selected directions and next production owner;
• if the user asks for more versions of selected ads, use the mood-board Remix action and append new results to the same board rather than creating a separate review wall.

Files

• assets/image-ad-library/: checked-in reusable ad library and registry.
• assets/image-ad-library/packs/diverse-image-ad-archetypes.json: default 25-family starter ad pack.
• assets/image-ad-library/packs/digital-product-core-ad-prompts.json: 12-route screen-first ad pack.
• scripts/build_ads_explorer.py: wrapper around the shared offer explorer prompt-pack engine.