Skip to content

gooseworks-ai/goose-skills / skills/marketing/plan-campaign

plan-campaign

Verified

Campaign planner: turns a product, offer, goal, budget and date into a plan the marketer can approve and a designer can start from today, with every channel kept or dropped for a reason. Zero paid is a valid answer; it never spends. Use when the ask names a campaign, a launch, drop or season push, a budget or a date: I have this budget and a launch date — what do I run, where?; plan my launch, drop or Black Friday push; plan a campaign; how should I split $5k; plan next month's campaign; our CAC blew up — replan. Not for a broad "grow my brand", "get more customers" or "where do I start" with no campaign, budget or date — use where-to-start; not for pausing ads on a sold-out product — use run-paid-campaigns; not for an email-only send plan — use email-lifecycle; not for reading this week's numbers — use analyze-performance.

references/intake.md

View on GitHub
# Intake — look first, then ask like a person

Load at Before you start, steps 2–3. A good marketer reads everything they can first, then asks three or four plain questions — never a form, never a technical question.

## 1. Look first (free, before the first message)

| Look at | It answers | Don't ask about it if found |
|---|---|---|
| `brand_knowledge_browse` (when `brand_knowledge_*` tools are listed) and its chapters 2, 3, 5, 15, 18, 19, 20, 23; else `brand_read` and `knowledge_search` | rules, products, what sells, who buys, approvals and budget, how results are counted, what's coming, targets and past tests, what is working | anything the brain already holds — items waiting for a yes are used, labelled as waiting |
| the last plan record and the run records since it (`references/run-memory.md`) | how last time's legs and tests did, money moves and their results | "how did last month go?" — say what the records show |
| `brand_read` products (with prices for the hero products) | what's for sale, plans and prices | which product exists, what it costs |
| `brand_read` onboarding | who makes the ads, who runs them, platforms used, spend band, stated goals | "who makes the ads?", "which platforms?" |
| `brand_read` learnings (full) + `knowledge_search` for the task | past results, offers that ran, rules | "what worked before?" |
| `meta_read` pages + performance | is Meta connected; recent spend and results on ads GooseWorks made | "do you run Meta ads?" |
| `social_account_list` | which organic accounts exist | "do you post on Instagram/TikTok?" |
| `campaign_read` + `ads_creative_read` (approvals view) | campaigns already planned, offers and audiences used, creative waiting | duplicates; "is this the Halloween one?" — say what you see instead |
| `lifecycle/program.md`, `lifecycle/calendar/<YYYY-MM>.md` | email/SMS list size, flows, sends already planned, standing offer | "do you have an email list?" |
| `reports/performance/definitions.md` + newest weekly read | margin, the real number (store / CRM), tracking state, recent results | margin, "is tracking set up?" |
| the last approved plan's money and status sections | AOV, margin, list size, founder hours given last time | anything already stated |
| `research/market/positioning.md` or product-specific `research/market/positioning-<product>.md`, plus the dated positioning report it cites | the owner-selected launch direction and messaging; unselected options are proposals only | do not ask the owner to choose again when the selection is recorded |
| research files (`gooseworks:file_read`, else `knowledge_search` documents): the market read and `demand.md`, buyer groups and the customer-words bank, the competitor report, `angle-bank.md`, `landscape.md`; `competitor_read`; `creators/roster.csv` when creators may be a leg | who buys, why, what rivals run, which creators exist | "who's your customer?" |
| the claims contract, through the `claims-gate` skill (`../../claims-gate/SKILL.md`) | which claims are usable for the hero product | "what can we say about it?" |
| **the brand's live site — required every run** (host web fetch / browser): home and pricing pages (the address is in the brand summary) | live offer, prices and current public positioning; confirms current commercial facts | "is there a sale on?" — confirm instead of asking |
| an existing plan for the same occasion | what was decided before — **input to refresh, never the answer**; its numbers are re-checked against today's facts | "is this the same campaign?" — say what you found and that you'll refresh it |

On the `brand_read` path, read lean first, then `product_ids` for the hero product, then learnings in full in their own call — never every section at full detail in one call (it overflows and gets skipped).

When sources disagree about current commercial facts (e.g. saved prices vs live pricing), use the live site and say which you used. An explicitly owner-selected launch brief is the planned direction even when the current site has not changed: name that difference and carry the selected positioning into the campaign. Do not replace it with old public copy or silently select an unapproved option. Claims still pass the claims gate. If the host has no way to read web pages, say so in the first message and label prices "from saved notes, unconfirmed".

**Local mode** (no GooseWorks, marketing-guide §5): the look-first pass is the live site, the local folder (`brand-notes.md`, `claims.md` through the claims gate, the last plan in `plans/<slug>/` — step 1's existing-plan check — and run records) and any connectors the marketer has (read-only). Everything the brain would have answered becomes a question or a labelled assumption.

Margin is looked up (the measurement contract, the last plan); if it's not there, it is a labelled assumption marked "confirm first" in the plan — ask it only when the whole plan turns on it.

Tracking is never a technical question. If a weekly read shows how buyers are attributed, use it; if not and the goal is paid conversions, ask the one plain-words question above; the paid run checks the setup itself before any money moves. When Meta is a likely leg and the goal is sign-ups or sales, fold Meta into that same question: "When someone signs up, can you see what brought them — and can Meta see the sign-up too?" (Meta can only aim at sign-ups it sees; if not, the leg is priced as Traffic — `unit-economics.md` §3).

## 2. The first message

Shape: what you found (2–3 lines — the rest belongs in the plan), then at most four questions in this order, in plain sentences — never a numbered form with sub-bullets:

1. **Product and offer.** Propose from what you found and ask to confirm; for a one-product brand, state the product and ask only about the offer.
2. **Who it's for: new people, or people who already know you** (past customers, free or lapsed users, the list), and roughly how many could be reached. This decides whether the plan is a cold-acquisition plan or a win-back plan, so it comes before the money.
3. **Goal, dates, budget** — one sentence, skip what they've said.
4. **What the plan covers** — say it, don't ask a question: every channel that could work (ads, email, organic, creators, founder outreach), kept only if it fits; they can say "ads only".
5. **At most one more gap that changes the plan** (e.g. whether a discount is possible), with why it matters and what you'll assume. When the goal is sales or paid accounts and nothing on file shows how buyers are attributed, the plain-words question is: "When someone upgrades, can you see what brought them?" (with Meta folded in when it is a likely leg, §1).

Example — a fictional brand, Fernly (invoicing app for freelancers), after the look-first pass. Never copy a real brand's facts from an example; use only what the look-first pass found.

> I had a look first: Fernly is one product, on Free, Solo ($12) and Studio ($29) monthly plans, with no offer on the site right now. Your ad account isn't connected here, and the only past campaign was a spring awareness push with no results on file.
>
> Is this for the paid plans, at full price, or is there an offer? And who is it for — new freelancers, or people who already signed up and drifted off? Roughly how many of those could you email?
>
> What's the goal (say, new paying accounts), the dates and the budget? I'll look at every channel that could work — ads, email, posts, a personal note from you to past users — and keep only the ones that fit. Say if you only want ads.
>
> Last one: when someone upgrades, can you see what brought them? If not, I'll plan around sources we can count, like tagged links and codes.

What the example does **not** ask: funnel rates, churn, margin, who makes the ads, pixels or link tags, platforms — those come from the look-first pass, or become labelled assumptions ("I've used a typical 3–5% of free users becoming paid; replace it when you know yours").

## 3. After the answer

- Only product, audience, goal, budget and date block planning. Everything else unanswered becomes a labelled assumption in "What we couldn't check".
- A follow-up question is allowed only when the answer changes the plan's shape (e.g. "1,000+ past users" turns a cold-ads plan into a win-back plan) — ask it in one line, with the plan you'll make either way.
- If the marketer pushes back on a question ("I don't know", "what kind of question is this?"), drop it, use a labelled default, and keep going.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 322kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 198kUpdated Claude CodeCodexVerifiedInstall