content-engine

name	content-engine
description	Use to turn NoblePort project activity into marketing — converting photos, videos, and daily logs into Facebook posts, LinkedIn articles, website case studies, and reel scripts. Use after a milestone, completed job, or site visit to generate publish-ready drafts. Outputs are drafts; publishing client work requires consent.

Content Engine Skill

Purpose

Make content a byproduct of operations: one field activity → 5–10 downstream marketing/sales assets, with no extra work for the field team.

When to use

• Photos/video/logs from a milestone or completed job are available.
• A case study, social post, or reel is needed from real project work.
• The marketing cadence needs feeding from active jobs.

When NOT to use

• Publishing without client consent on identifiable client work (consent-gated).
• Customer-care communications → 08-trust-pipeline.

Inputs

• Project artifact: photos, video, daily log, or completed-job record — with project name, service line, location, summary, highlights, metrics, and the client consent flag.

Workflow

1. Capture the artifact and its fields (don't invent missing ones).
2. Run the playbook for that artifact type to fan out the right channels.
3. Render drafts (post/article/case study/reel script) from artifact fields; missing fields become [[provide: …]] markers, not fabricated copy.
4. Hold for approval: every asset is DRAFT, or BLOCKED if it's a consent-gated channel without consent on file.
5. Human approves → asset advances; only then is it publish-eligible.

Outputs

• Facebook posts · LinkedIn articles · website case studies · reel scripts
• Portfolio entries · before/after · testimonial requests · weekly summaries

System integration

• Directly backed by the Journey Agent (backend/agents/journey.py, backend/journey/).
• API: /api/journey/process-artifact, /playbook/{type}, /assets, /assets/{id}/approve, /story-engine, /flywheel.
• Model: backend/models/journey_asset.py (hash-chained asset ledger).

Guardrails

• Draft, never auto-publish. The engine cannot publish; a human approves.
• Consent gate on any externally published asset about an identifiable client project — BLOCKED until consent is recorded.
• No fabricated facts — unknown metrics/details are surfaced as gaps.

Success criteria

• Each qualifying artifact yields 5–10 on-brand drafts (the leverage target).
• No client work is publishable without recorded consent.
• Every published asset traces to its source artifact and approver.