| name | marketing |
| description | One command to run BuildOS marketing ops from a durable queue instead of memory. Reads docs/marketing/ops/queue.json + cadence.json, runs the deterministic status engine (scripts/marketing/ops/status.mjs), and routes into the four pipeline stages — QUEUE/STATUS, ASSETS, PUBLISH, ENGAGE/LEARN — reusing the anti-feed, draft-anti-feed-blog, ideate/moodboard, and platform warmup skills. Triggers on "/marketing", "marketing status", "what's due to post", "marketing backlog", "spin up what I need to post", "log metrics for X", "what marketing is overdue". |
Marketing Ops Pipeline
The single entry point for running BuildOS marketing as a pipeline off a queue, not a set of ad-hoc sessions. The queue is the memory between sessions — you always pick up exactly where you left off.
Full design + schema: docs/marketing/ops/README.md (read once for context; don't re-read every run).
The prime directive
ALWAYS run the status engine first, every single invocation, before doing anything else:
node scripts/marketing/ops/status.mjs --today=<today>
Pass --today=YYYY-MM-DD using the real current date from the environment context (the script defaults to system time, but pin it so the report is deterministic and matches what the user sees). This prints the 🔴 overdue / 🟡 due / 🟢 ready-to-post / 🔵 next / 🖼️ asset-gaps / 📦 backlog picture. Everything below is a response to what that report shows.
Never hand-compute cadence or "what's overdue" yourself — that math is the engine's job (rerouted to a deterministic tool on purpose). Your job is to read its output and act.
Routing (args)
| Invocation | Do this |
|---|
/marketing (no args) | Run the engine, show the report, then recommend the single highest-leverage next action (top 🔴, else top 🟢, else 🔵). Offer to start it. |
/marketing backlog | Backlog synthesis — see references/backlog-synthesis.md. |
/marketing assets [id] | Stage ② — see references/stage-assets.md. |
/marketing publish [id] | Stage ③ — see references/stage-publish.md. |
/marketing post [id] [platform] | The posting sub-step of stage ③ — drive the platform skill, then mark the deliverable posted. |
/marketing engage / /marketing metrics | Stage ④ — see references/stage-engage.md. |
/marketing add "<idea>" | Add a new queue item (idea). See references/queue-editing.md. |
/marketing done <id> <deliverable> [url] | Mark a deliverable posted (with URL) or a blog published. See references/queue-editing.md. |
If args are ambiguous, run the no-arg flow (status + recommend) and ask which thread to pull.
The four stages (what each reuses — do NOT rebuild)
- ① QUEUE / STATUS — the engine +
queue.json. This skill owns it.
- ② ASSETS —
/ideate and /moodboard commands + the Real Media Policy (docs/marketing/brand/BUILDOS_REAL_MEDIA_POLICY.md — zero AI images of product/founder; real screenshots/recordings/Inkprint cards only). Resolves each deliverable's asset_needs → assets.
- ③ PUBLISH — drafting reuses the
anti-feed skill (publish kit, menu option 2) and draft-anti-feed-blog (the blog itself); posting reuses the twitter / linkedin / instagram / reddit browser skills. Voice comes from the brand guide + per-platform voice quick-refs (see stage-publish).
- ④ ENGAGE / LEARN — pull post metrics via the platform browser skills into each item's
metrics[], then synthesize what's working to steer the next posts.
The core loop (no-arg flow)
- Run the engine (prime directive).
- Read the report. Identify the one highest-leverage move:
- Any 🔴 overdue → the oldest/most-blocking one.
- Else any 🟢 ready-to-post → the one with least window left.
- Else 🔵 next blog due → draft it.
- State the recommendation in one line + why, then offer to run it (respect DJ's vision-first + decision-brief style — don't dump the whole report unless asked; lead with the recommendation).
- When the user does an action (drafts a kit, posts a lane, generates an asset), update
queue.json immediately (see references/queue-editing.md) and re-run the engine so the new picture is visible. The queue must never drift from reality.
Golden rules
- The engine output is the source of truth for timing;
queue.json is the source of truth for state. Keep them honest — patch the queue the moment something changes.
- Reuse the existing skills for content generation and posting. This skill orchestrates; it does not re-implement drafting, kit-building, or browser posting.
- Respect the Real Media Policy in every asset decision.
- One recommendation at a time. Marketing debt is paid down one overdue item at a time, not in a batch panic.
- After any queue edit, re-run the engine and show the delta.
