Skip to content

gooseworks-ai/goose-skills / skills/marketing/make-static-ads

make-static-ads

Verified

Make new static image ads, a fresh batch, more ads, a winner's variations, copy variants or a rejected ad fix. Reads the brand, its customers and the ads it already runs, turns real product facts and customer situations into distinct visual ideas, shows a concept board, writes each render prompt, quotes, renders and reviews through GooseWorks. Also handles requested branded header/social graphics. Use for static ad creation rather than goose-ads. Not for video, product photos or launching campaigns.

references/engine-notes.md

View on GitHub
# Engine notes — how the GooseWorks ad tools behave

The live tool answers are the truth for parameters. These notes cover what the tool descriptions don't say.

- **`ads_render` is thin.** It sends the agent's `prompt` and `image_urls` (public https, in order) to the image model word for word and does nothing else. The agent writes the whole prompt (`prompts.md`) and judges the result (the `make-static-ads-render-review` skill (`../../make-static-ads-render-review/SKILL.md`)). Every retry is a paid render.
- **What an ad is based on.** `ads_render` takes `template_id` or `reference_image_url` (one, not both) and records it on the creative so the app shows it; neither attaches an image. The dry run returns `template: {id, title}` when the id resolves, and refuses an unknown id before any quote.
- **Read small, read once.** Tool answers can be huge (a full kit is 50–150K characters; an unfiltered creatives list is over 1 MB). Use these calls, and on a shell host save the big ones to a file once per batch and read fields from the file (`jq`) instead of reading the answer again:

| Need | Call | Size |
|---|---|---|
| Brand, products, competitors, `must`/`dont` learnings | `brand_read` `sections: ["summary","kit","products","competitors","learnings"]` (lean, the default) | small |
| One product's full record | `brand_read` `product_ids: [<id>]` | small |
| Kit and learnings in full, for prompts (once per batch) | `brand_read` `sections: ["kit","learnings"]`, `detail: "full"` → save to `kit.json` | very large: save once, read every field used by this batch |
| Brand facts and claims | `brand_knowledge_browse` once, then `brand_knowledge_search` with `source_types` and `product_id`, `limit` ≤ 25 | medium |
| Ads this playbook made before | `ads_creative_read` `tags: ["made-by:make-static-ads"]`, `limit: 20` | small |
| Other ads made in GooseWorks (in the app, or before this tag existed), to reach the last 10 | the brand's batch files first; then `ads_creative_read` `approved_only: true`, `limit: 10`, saved to a file on a shell host; no shell → skip this read: the history is the batch files plus the tagged list above, and the Read line says so. Open single ones with `creative_id`. Never page through the list | very large: video projects ride along on every row (10 rows were over 700K characters in a test) |
| Own and competitor imported ads | `ads_template_read` `mode: "mine"` / `"competitor"`, `filters.format: "image"`, `limit: 20` | medium |
| One template | `ads_template_read` `template_id` | small |

Never `detail: "full"` on every section at once, never the `sources` section (saved web passages; it alone can pass 100K characters), and never a creatives list without `tags`, `status`, `batch_id` or `approved_only`. An answer over about 50K characters: save it to a file once and read fields from the file.
- **Polling.** `job_get` (`kind: "ads_batch"`) is done when every creative's `pending` is 0. Each answer repeats the full prompt, so poll every 30–60 seconds, or check the render's image URL instead.
- **Edits are unquoted.** `regenerate`, `precision_edit` and `resize` return no dry-run price, so they are not used. Every remake, text fix, extra placement and copy variant is a written, quoted `ads_render` call.
- **Campaign records are input only.** `campaign_read` gives an approved campaign's brief and concepts (use them only when the brief's Status line says "Approved"); this playbook never generates through the campaign tools.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall