gooseworks-ai/goose-skills / skills/shared/meta-ad-manager
meta-ad-manager
UnverifiedFront door and loop owner of a Meta (Facebook/Instagram) ad harness. Reads the brand's ads/ docs, works out which campaign the user means and which stage it is at, hands off to the stage skill that runs it (intake, strategy, creative, launch, checks, fixes, answers), and writes every stage change back to the docs. Use it when someone wants to run ads on Meta, comes back to ads they started earlier, a stage skill hands back, or a scheduled check fires.
contract/RULES.md
View on GitHub# Meta ad harness: file rules
Every harness skill follows these rules when it reads or writes the brand's ad docs.
- **Where:** one `ads/` folder per brand. The mode's adapter (in the `meta-ad-manager` skill's
`references/`) says where that folder is.
- **Who shares it:** every run for the brand (chat, scheduled checks, reports) reads the same
files.
- **How:** whole-file reads and writes with the host's file tools. Never shell edits
(`echo >`, `sed -i`).
## 1. Where the files live
| Mode | Where `ads/` is |
|---|---|
| **GooseWorks** | the brand's coworker workspace. See the GooseWorks adapter for the exact calls, both inside the coworker and from an outside host |
| **Direct Meta** or **Planning only** | a local `ads/` folder in the working directory the user chose for this brand. One folder per brand |
**Ids outside GooseWorks.** `ads/README.md` and `ads/brand.md` need `brand_id` and
`coworker_agent_id` in their frontmatter. With no GooseWorks brand record, write `local` for
both. Never leave a template placeholder and never write `none`.
**Evidence pointers outside GooseWorks.** A Meta read is `meta:<object id> <since>..<until>`
(for example `meta:act_123 2026-09-17..2026-09-23`). GooseWorks uses `report:`, `push:` and
`finding:` ids that its tools return.
## 2. How to read and write
- Read a file whole, change it, write it whole. Use the host's own file tools, or the calls the
GooseWorks adapter names.
- Never hard-code an absolute path in the docs. The folder can move between machines or sandboxes.
- If two routes can reach the same files (for example a sandbox and an outside host), follow the
adapter's note on caching before you write from one and read from the other.
<!-- shared:harness-contract start -->
<!-- This block is maintained in ONE place and copied byte-for-byte into both the
maintainers' source copy and the published goose-skills copy. Edit one, copy
it to the other, and run the parity script before shipping either. -->
## 3. Read order
1. `ads/README.md` first, always.
2. Then only the campaign folder in play: `ads/campaigns/<slug>/state.md`, plus `strategy.md`
or `decisions.md` when the stage needs them.
3. `ads/brand.md` body only for intake, strategy and deep checks, or to find a business fact
before asking the user for it. Its frontmatter (`familiarity`, `business_type`) may be read on
any turn that talks to the user, to set the tone.
Do not list or read other campaigns "for context".
## 4. How to write
- **Whole files.** Read the file, change it, write all of it back. A write replaces the whole
file.
- **Keep the index in step.** When a campaign's stage changes, update its row in
`ads/README.md` in the same turn. The index Stage must equal `state.md` `stage`.
- **`decisions.md` is append-only.** New entries go at the end. You may fill in one earlier
line, `- Outcome: pending`, once the result is known. Nothing else may be edited or removed.
- **Follow the templates** in `templates/`. Drop the `<!-- -->` guidance comments when you
write.
## 5. Numbers: decisions vs observations
- **Goose's own decisions carry plain numbers**: "$30 per day for 14 days", "3 ads from 9",
"CAC target $45 (per the owner)". No date is needed. They are plans, not facts about Meta.
- **Anything read from a tool is an observation.** It goes on its own `- Observed:` line with
its data window, sync time and source:
```
- Observed: cost per subscriber $41 across 2 ads [window 2026-09-23..2026-09-23; synced 2026-09-24T09:00Z; source report:rep_0924]
```
Never call an observation "current", "now", "today" or "latest". It is a snapshot.
- **Meta facts are not stored as truth.** Read spend, delivery status and rejections from the
tools when you need them. The docs record what Goose decided, why, and a pointer to the
evidence (`report:<id>`, `push:<id>`, `finding:<id>`).
## 6. Few files, small files
- `ads/` holds exactly `README.md`, `brand.md` and `campaigns/<slug>/{strategy,state,decisions}.md`,
plus an optional `decisions-archive.md`. Nothing else: no per-poll, per-finding or per-report
files. Those stay in the database, and the docs point at them.
- Each file is at most 8 KB. `decisions.md` is capped at 64 KB. From 48 KB on, move closed
entries verbatim to `decisions-archive.md`.
- Campaigns are never deleted. Set `stage: ended` instead.
## Checking a folder
```bash
python3 validate.py path/to/ads # structure and rules
python3 validate.py path/to/ads --baseline old/ads # also enforce append-only
```
<!-- shared:harness-contract end -->
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 198k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |