Skip to content

gooseworks-ai/goose-skills / skills/shared/meta-ad-manager

meta-ad-manager

Unverified

Front 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.

references/gooseworks-adapter.md

View on GitHub
# GooseWorks adapter

## Summary

Use this mode when the GooseWorks tools are callable: inside the GooseWorks coworker (chat,
messaging channels, scheduled automations) or from any agent host with the GooseWorks MCP. The
`ads/` docs live in the brand's coworker workspace, so chat, scheduled checks and outside hosts all
see the same files.

## The `ads/` docs

**Inside the coworker** (chat, a messaging turn, a scheduled run): use the built-in `Read`,
`Write` and `Edit` tools with a path **relative to the working directory**, which is the workspace
root. The coworker has no `file_*` tools there.

```json
{ "tool": "Read",  "args": { "file_path": "ads/README.md" } }
{ "tool": "Write", "args": { "file_path": "ads/campaigns/<slug>/state.md", "content": "<whole file>" } }
```

Never write to an absolute path such as `/workspace/…`: that is the sandbox's local disk, which
is lost when the sandbox stops and which no other run can see.

**From an outside host** (Claude Code, Cursor, any MCP client): call `brand_read` without `brand_id`, take the
brand's `coworker_agent_id`, and pass it as `scope` on **every** file call. Without `scope` the
tools fall back to the user's default agent, which is usually not the brand's.

```json
{ "tool": "brand_read", "args": {} }
{ "tool": "file_read",  "args": { "path": "ads/README.md",
    "scope": { "type": "agent", "agent_id": "<brand.coworker_agent_id>" } } }
{ "tool": "file_write", "args": { "path": "ads/README.md", "content": "<whole file>",
    "scope": { "type": "agent", "agent_id": "<brand.coworker_agent_id>" } } }
```

**Ids.** `brand_id` is the GooseWorks brand id and `coworker_agent_id` its coworker agent, both
from `brand_read` without `brand_id`.

## Meta tools from an outside host

The launch, pause and proposal tools take **`brand_id`** on every call from an outside host (the
GooseWorks brand id from `brand_read`, not the coworker agent id). Inside the coworker they take no
brand: the session is already bound to one. Everything else about them is the same, and so are
the gates: they act as the signed-in user, on a brand in the user's own organisation.

```json
{ "tool": "meta_read",            "args": { "brand_id": "<brand.id>", "view": "pages" } }
{ "tool": "prepare_meta_ad_push", "args": { "brand_id": "<brand.id>", "render_ids": ["<render id>"],
    "mode": "NEW_CAMPAIGN", "page_id": "<id from meta_read view pages>", "...": "..." } }
{ "tool": "pause_meta_ad_push",   "args": { "brand_id": "<brand.id>", "push_id": "<push id>",
    "meta_ad_ids": ["<meta ad id>"], "user_instruction": "<the user's words>" } }
{ "tool": "propose_meta_budget_change", "args": { "brand_id": "<brand.id>", "push_id": "<push id>",
    "new_budget_minor_units": 5000, "reason": "<why, with the evidence>" } }
```

No tool approves, executes or activates anything. `prepare_meta_ad_push` and the two `propose_*`
tools return an `approval_url`: the user approves there, in the GooseWorks app. Give them the link
and never say it is done.

**Caching.** The coworker's file mount caches file info for up to two minutes. After an outside
`file_write`, a coworker run can see a stale or missing file for that long. Do not write a harness
file from outside and immediately start a run that reads it; wait, or put the change in the
message instead.

## Surfaces

- Web chat and messaging channels both pass Step 0. The per-turn context says which one it is.
- Scheduled checks are GooseWorks automations. They have no user message; the prompt names the
  check (daily quick check, weekly deep check).

## Stage → tools

| Stage | Tools |
|---|---|
| intake | `ad-management-intake` (its GooseWorks adapter): `brand_knowledge_browse` / `brand_read`, `brand_setup_status`, `campaign_read` / `campaign_upsert` for the brief |
| connect | the in-app connection card (`render_meta_connection` with only an intent inside the coworker). Never ask for or repeat a token |
| strategy | `ads-north-star-strategy` (its GooseWorks adapter) |
| create, review | concepts via `add_campaign_concept`, then `request_campaign_generation({campaign_id, planned_ads})`, which generates three times the plan and waits in the approval queue until the user approves (the Generate button or `ads_approval_decide`). Tell the user where to approve. A one-off "make me an ad" outside a plan uses `ads_generate` with `dry_run: true` first (the cost quote; it quotes template sources only), then the same call without `dry_run` after the user's yes. `ads_creative_read` (`approved_only: true`) fetches the picks |
| launch | `launch-meta-ad-campaign` (its GooseWorks adapter): `meta_read` view `pages`, `prepare_meta_ad_push`, the approval page, then `meta_read` view `push_status`. No `refresh_meta_sync` after a launch: the push status already lists every object created; refresh only when an object you know exists is missing from reads |
| quick check | the daily Meta ads report automation and its watch findings |
| deep check | `meta-ads-analyzer` over `query_meta_insights` complete days |
| fix and adjust | `pause_meta_ad_push` (a whole push or one ad), `revert_meta_ad_push`, `propose_meta_budget_change`, `propose_meta_creative_replacement`, then `meta_read` with `view: "change_status"` and the `change_id` after the user approves on the page |
| answers | `answer-ads-questions` (its GooseWorks adapter) |

`render_meta_connection` works inside the coworker only: from an outside host, send the user to
connect Meta in the GooseWorks app. The Meta read tools (`meta_status_get`, `list_meta_entities`,
`query_meta_insights`, `get_meta_ad_context`) may not be listed on an outside connection yet. When a
tool is not listed, say that step needs the brand's coworker in GooseWorks, and do not guess.

## Evidence pointers

In `decisions.md`: `report:<id>` from the daily report, `push:<id>` from `prepare_meta_ad_push`,
`finding:<id>` only when a daily report names that finding id. There is no tool that lists watch
findings, so never make one up.

## Whose ads are Goose's

GooseWorks records every launch as a push. The launch record for `answer-ads-questions` rule 5 is
the push list in `state.md` (`meta_push_ids`), and both pause and revert exist.
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: 472kUpdated Claude CodeCodexVerifiedInstall
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: 198kUpdated Claude CodeCodexVerifiedInstall
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: 198kUpdated Claude CodeCodexUnverifiedInstall
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: 198kUpdated Claude CodeCodexVerifiedInstall