plan

name	plan
description	Build or refresh the planning layer — the company plan, the CRO-level strategy, the concrete outcomes, and the who's-who. Runs the gtm-strategist through an interview loop and grounds the plan in the workspace graph. Usage: /plan [layer-1|layer-2|layer-3|people | a free-form focus]

/plan

Build or deepen the planning layer. This is where the business's goals, revenue strategy, and concrete outcomes become living documents the rest of the harness executes against.

$ARGUMENTS

If an argument names a layer (layer-1/company, layer-2/strategy, layer-3/outcomes, people) or a focus area, scope the run to that. Otherwise do a full refresh across all three layers and the cast.

---

Step 1 — Read what exists

Read the current state of the plan before changing it:

• planning/COMPANY_PLAN.md, planning/STRATEGY.md, planning/OUTCOMES.md
• workspace/PEOPLE.md

Note what's already settled and what's marked TODO: or ASSUMED: — those open items are this run's priority. Preserve what's good; deepen what's thin. Don't rewrite settled sections without reason.

---

Step 2 — Ground in the workspace

Spawn the data-analyst subagent to refresh the facts the plan should rest on:

Pull a current workspace snapshot for planning: people & roles, agent/skill configuration coverage,
pipeline shape (and whether it's maintained), and activity/coverage gaps. Follow your output format.
Flag anything stale or permission-gated.

This keeps the plan honest — forecasts tied to real pipeline, outcomes tied to real coverage gaps, owners tied to real people.

---

Step 3 — Run the strategist's interview loop

Spawn the gtm-strategist with the data-analyst's snapshot and the existing plan:

Refresh the planning layer{, scoped to {layer/focus} if specified}. Here is the current workspace
snapshot:

{paste data-analyst output}

And the current plan documents are in planning/ and workspace/PEOPLE.md. Resolve the open TODOs,
deepen thin sections, and keep the three layers coherent — every Layer-3 outcome must ladder up to a
Layer-2 strategy and a Layer-1 goal, name an owner from PEOPLE.md, and (where it implies a workspace
change) state what it becomes. Run a focused interview — one question at a time, each with your best
guess. Return the planning update summary.

Run the interview in the main thread so the operator answers in real time. One question at a time, each carrying your best guess so they can correct rather than compose.

---

Step 4 — Write the documents

The strategist writes planning/*.md and workspace/PEOPLE.md directly, filling the existing templates. After it returns:

• Show the operator a concise diff of what changed across the four documents.
• List any TODOs that remain open and the specific question blocking each.
• Confirm coherence: every outcome has an owner who's in the cast, and every layer ladders up.

---

Step 5 — Point to execution

Close with the bridge to implementation:

## Plan refreshed

- **Layer 1 (Company Plan):** {one line on the state}
- **Layer 2 (Strategy):** {one line}
- **Layer 3 (Outcomes):** {N} outcomes, {N} owned, {N} that imply a workspace change
- **Open TODOs:** {N}

### Next
- `/audit`      → see which outcomes the workspace already delivers and where the gaps are
- `/sync-pages` → publish the updated plan to Day AI Pages
- `/start`      → take stock of initiatives; if a strategy shift here implies a bounded, owned effort, capture it as a new `initiatives/<slug>.md`

---

Notes

• Plan only — no workspace writes. /plan reads and lists the workspace (via the data-analyst and strategist) but never invites, edits agents, or creates skills. That's /implement.
• Keep the plan decision-ready, not exhaustive. A tight plan an implementor can act on beats a sprawling one.
• If the workspace snapshot contradicts the plan (e.g. the forecast assumes a pipeline that isn't maintained), surface the contradiction in the interview — don't quietly write around it.
• Plan vs. initiative. /plan keeps the standing intent and the fine-grained outcomes current. When the plan implies a bounded, owned effort with a deadline and a verifiable definition of done ("get the SE team live on Day AI by Q3"), that's an initiative, not an outcome — note it so it can be captured in initiatives/ and driven by /start. See initiatives/README.md.