6.**`progress.md` exists, `phase: finalized`** → plan is done. **Do not silently overwrite.** Ask the user: *"This plan is finalized (v{N}). Want to (a) revise it as v{N+1}, (b) start a fresh plan in a new folder, or (c) re-open a specific section?"*
Update `phase` and `last_updated` whenever state changes.
- For the team's current marketing owner (if there is one): is the shape π-shaped (two deep skill sets), T-shaped (one deep, broad), or tactical-only? See `team-and-agency-model.md` for the framework that informs Section 11 RACI and the first-hire recommendation in Section 9.
- Blended CAC if known (including salaries, content costs, tools, retainers — not just paid ad spend). If unknown, flag as the top Section 13 open decision — every revenue projection depends on it.
- ARPC, annual retention rate (or churn rate), so the budget math in `budget-planning.md` can be applied to Section 8 (Revenue) and Section 10 (12-month outlook).
- Acquisition: organic SEO, paid search, paid social, content, social, partnerships, events, PR, ambassadors, etc. — for each, status (live / paused / never tried)
- **From rich materials.** When the team has shared decks, prior content audits, an existing brand voice doc, recent positioning work, or a kickoff call transcript — score from those. Mark "scored from materials" in the section heading.
- **From a separately scored audit.** If the team already has a scored current-state assessment (in any format), ingest those numbers directly. Don't redo the work.
Either way, the output is the scored 17-row table that becomes Section 3 of the plan, followed by a 2–4 sentence "shape interpretation" calling out where strengths and gaps cluster.
[17 section scores using `references/current-state-rubric.md`. If a prior scored audit exists, paste those scores. Otherwise mark "scored from materials."]
**Goal:** Walk through all 13 sections of the plan template (`references/plan-template.md`), drafting each, getting user confirmation, saving as you go.
Section 1 (Executive Summary) is drafted **last** because it depends on every other section's conclusions. Walk Sections 2 → 13 in numeric order, then synthesize Section 1 from the others. The final compiled `final_plan.md` is always presented in canonical order 1 → 13.
4. Save the confirmed text to `sections/01.md` ... `sections/13.md` (one file per section, zero-padded for sort order). This is the canonical persisted artifact — recovery depends on it.
**Section 1 (Executive summary)** is synthesized from Sections 2–13 after they're all approved. Draft it last; present it first in the output document.
**Section 3 (Current state)** uses the embedded 17-section rubric in `references/current-state-rubric.md`. If a prior scored audit exists, paste those scores in. If not, score from available materials.
**Sections 4–8 (AARRR)** each follow the same internal structure: current state, the plan (numbered moves), 90-day moves, 12-month outlook, skills + tools. Don't skip the skills + tools sub-section — it's what makes the plan operationally honest.
**Section 11 (Marketing operations stack)** is auto-generatable from `references/ops-stack-mapping.md` plus the specific moves named in Sections 4–8.
**Section 12 (Idea bank)** is auto-generatable from `references/idea-cross-reference.md` plus client-specific filters (skip ideas that conflict with brand voice; status moves based on funding-stage timing).
**Section 13** lives at the end. Open decisions should be ranked by impact. Appendix should reference only files the team can access (warn about machine-local paths).
### Step 2.4 — Brand voice consistency
If the client has documented brand voice rules (captured in research.md / Section 2), every section must respect them. Common voice constraints:
- Vocabulary rules (YES / NO lists)
- CTA rules (e.g., "never pressure")
- Initiatory vs. explanatory framing
- Tone (e.g., authoritative-yet-accessible, intimate-yet-professional)
If a section's draft violates the brand voice, redo it before showing it to the user.
Set `phase: finalize` in `progress.md` before starting. Concatenate `sections/01.md` through `sections/13.md` into `final_plan.md` (canonical order 1 → 13, regardless of drafting order). Add:
- **Cross-reference check** — every marketing-ideas number (e.g., "idea #17") matches the actual idea in `references/idea-cross-reference.md`. Every related-skill mention either exists in the `marketingskills` repo or is documented as an external dependency (see ops-stack-mapping note on cross-marketplace skills).
- **MCP/API check** — every tool mentioned in Section 11 actually exists in the user's stack (per research.md intake) OR is flagged as "future / not yet wired."
- **Path check** — no machine-specific paths (`/Users/...`, `/home/...`) in the output. Replace with descriptive references.
- **Voice check** — final read against brand voice rules. Flag and fix violations.
- **Open-decisions check** — every "TBD" or unanswered question from intake is listed in Section 13's open decisions, not hidden in the body.
- **Acknowledge check** — every item from "already done" in research.md is acknowledged somewhere in the plan.
### Step 3.3 — Print
Output `final_plan.md` to the plan folder. Print a summary to chat:
> *"Marketing Plan v1 saved to `~/marketing-plans/{client-slug}/final_plan.md`. ~X,XXX words across 13 sections. Ready to paste into Notion or share with the team."*
### Step 3.4 — Publish (optional)
Ask the user:
> *"Want me to publish this to a shared GitHub repo so the team can access it? If yes, what's the target repo and path (e.g., `{client-org}/{client-context}/marketing/plan.md`)?"*
If yes:
- Clone (or assume cloned) target repo
- Check out a feature branch or push direct to main per user's preference
- Copy `final_plan.md` to the target path
- Adjust the appendix to use repo-relative paths (not machine paths)
Set `phase: finalized` in `progress.md` and stamp `last_updated`. This is the terminal state and prevents future `/marketing-plan` invocations from silently overwriting the plan (see Step 1.1.2 case 6).
- **Skipping intake.** A plan written without proper intake is generic and won't survive contact with the founder. Always do the full ten-topic intake unless the user explicitly waives it.
- **Pretending data exists.** If you can't confirm a number (current MRR, retention rate, etc.), don't guess. Mark it `[TBD — to confirm with team]` in the plan and add to open decisions.
- **Ignoring the brand voice.** If the client has a strong voice (most do), every section must respect it. Read the voice rules before drafting any copy-adjacent text.
- **Padding the idea bank.** Section 12 is comprehensive only if it includes the skip list with reasons. Don't pad with ideas that clearly don't fit just to hit the 139.
- **Glossing over uncomfortable metrics.** If churn is high or activation is low, name it in Current State. Founders read past sugar-coating.
- **Forgetting funding-stage logic.** If the client is mid-raise, the plan must explain what changes when the round closes. Skipping this turns a plan into a wish-list.
