marketing-setup

Files (2)

• SKILL.md
• writing-rules.md

SKILL.md (first lines)

# Marketing Setup

Use this skill before content production begins.

## Boundary

Marketing setup prepares brands, profiles, accounts, guides, and content systems.

Do not manage:

- leads
- contacts
- sales accounts
- deals
- outreach cadences
- CRM follow-up

If setup reveals sales-ready context, note it for handoff to the user's CRM instead of tracking it here.

## Required Sequence

Do these steps in order:

1. Ask setup questions.
2. Research competitors, audience language, SERPs, and content patterns.
3. Build the visual brand guide.
4. List or connect necessary marketing accounts.
5. Build or refresh public profiles.
6. Start the posting/content pipeline.

Do not skip to posting until steps 1-5 are complete or explicitly waived by the user.

## Workspace Folders (create on first init)

A fresh clone ships with **no private workspace folders** — the repo only contains the app, skills, templates, and anonymous examples under `data.example/`. On first init, create the user's local, gitignored working folders (they never get committed):

- `data/`: the live workspace the app reads and writes. Seed it by copying the shape of `data.example/` (tenant-first folders, plus tenants, accounts, guides, research, calendar). Per-tenant init state lives at `data/<tenant>/setup/init.json`.
- `strategy/` — optional. Long-form working strategy documents the user maintains by hand (positioning, brand plans, content calendars). One markdown file per topic.
- `docs/` — optional. Architecture notes, design specs, and implementation plans for the user's own workspace.

Confirm `.gitignore` already excludes `data/`, `strategy/`, and `docs/` before writing anything into them, so private context never lands in a commit. Create only the folders the user actually needs; `data/` is the only required one.

## Intake Questions

Ask one batch of concise questions:

- What tenants or brands should this workspace manage?
- Which tenant is highest priority right now?
- What outcome matters most for each tenant?
- Which websites, blogs, and social accounts already exist?
- Which accounts need to be created or refreshed?
- What content assets already exist?
- Who are the main competitors or reference accounts?
- What should the visual style feel like?
- What topics, claims, or tactics should be avoided?
- What would make the first 30 days successful?

## Outputs

Create or update:

- `data/tenants/`
- `data/accounts/`
- `data/guides/`
- `data/research/`
- `data/calendar/`
- `data/<tenant>/content/`

Use `data.example/` only for anonymous public examples.

---

## 7-Stage Tenant Init Procedure

The Init wizard walks a tenant from zero to `readyToPost` through seven ordered stages. Each stage must be approved before the next one can begin. The dashboard wizard shows a copy-paste handoff prompt for agent stages in this exact format:

```
Run Init step "<stage-id>" (<Stage Label>) for <Tenant Name>
```

Paste that string into the chat to trigger the relevant agent work for that stage and tenant. The dashboard displays this string verbatim — the chat trigger must match it exactly.

Once a stage is saved (`in-review`), the wizard offers a second trigger alongside the Approve button, so the user can evaluate and refine the saved artifact in chat instead of approving blind:

```
Evaluate and refine Init step "<stage-id>" (<Stage Label>) for <Tenant Name>
```

When you receive this, read the saved artifact for that stage, assess it against the prior approved stages, and discuss or revise it with the user rather than treating it as final. The stage stays `in-review` until the user approves.

Status values: `not-started | in-progress | in-review | approved`

State is persisted at `data/<tenant>/setup/init.json`.

---

### Stage 1: `import-intake` — Import & intake

**Type:** Input (user fills; no agent handoff)

**Input artifact:** User supplies LinkedIn URL, website URL, existing brand assets (logos, PDFs, images), and free-text notes via the wizard intake form and asset uploader.

**Agent writes:** `data/<tenant>/setup/intake.md` (structured intake notes) and uploads to `data/<tenant>/setup/assets/`.

**Review gate:** User reviews the saved intake notes and asset manifest, then approves in the wizard.

**Loop-back:** If information is missing or incorrect, the user edits the intake form and re-saves before approving.

---

### Stage 2: `icp` — ICP

**Type:** Input / agent-assisted

**Handoff prompt:** `Run Init step "icp" (ICP) for <Tenant Name>`

**Input artifact:** Approved intake (stage 1), user-supplied audience notes.

**Agent writes:** `data/<tenant>/setup/icp.md`: Ideal Customer Profile covering job titles, industries, pain points, goals, objections, and where the audience spends time online.

From esot321c/marketing-ops, MIT.

More from esot321c/marketing-ops