Dineway Content Brief

Dineway Content Brief

Execute one Brief Stage that the master Pipeline already began. Use only the accepted Research Result as evidence authority. Return compact editorial decisions; the server compiles the complete Brief from accepted Research and Stage Complete renders canonical Markdown and derives its receipt.

Required inputs

• Current Run, Brief Job, active Assignment/Attempt, and exact accepted Research Result ID/version.
• Accepted Research payload, plus its derived evidence.json and findings.md only as readable projections.
• Native collection/content identity, locale/market, byline, current schema fingerprint, Site Briefing, Site Context, applicable template, and public content inventory.

Stop if the accepted Research selection changed, required evidence expired, or native provenance does not match the derived Research projections. Do not rerun hidden research inside Brief.

Writer-ready Brief

1. Preserve Research topic, market, audience, objective, keyword metrics, intent, relevant questions, competitor gaps, and evidence lineage.
2. Choose exactly one primary target keyword. Secondary keywords preserve their Research metrics, intent, category, and Observation IDs. Record primary and secondary search intent explicitly.
3. Define audience role, expertise, pain point, funnel stage, voice, objective, unique angle, subjects, required topics, exclusions, and risky-claim guardrails.
4. Produce a single ordered outline with these required section types:
   • exactly one direct_answer;
   • exactly one at_a_glance;
   • one subject section per declared subject, grouped under subject_sections;
   • exactly one methodology;
   • optional faq using only Research questions marked included and includeInFaq=true; and
   • exactly one cta.
5. Give every section a stable ID, heading, level, format, required key points, evidence Observation IDs, question bindings, and word budget. Required key points must have Observation lineage. Section budgets must total within ten percent of the target word count.
6. Carry only approved Research questions. Rejected questions retain their reason in Research and cannot enter Brief, the outline, or FAQ.
7. Add verified internal-link candidates by stable public URL plus media purpose, alt-text intent, optional generation prompt, and whether each item is required.
8. Bind the Brief to the exact Research Result, collection, optional existing content ID, translation group, locale, byline, and schema fingerprint.

Never invent menu, hours, price, allergen, award, review, staff, medical, legal, or financial facts. Do not hide optional ideas among required scope.

Stage completion

Return only target/audience/voice selections, guardrails, internal-link/media choices, approved questions, the ordered outline, provenance, and selected persisted Observation IDs. Do not repeat Research metrics, locale, schema, CMS identity, or upstream bindings; the server hydrates them. Do not build an artifactRef, Result envelope, Markdown wrapper, hash, byte count, or acceptance write.

An authenticated human reviews the structured Brief and calls content_pipeline_stage_complete with briefApproval.confirmed=true. API-token and system actors cannot approve it. The transaction validates Research inheritance and outline completeness, renders:

• .dineway/content/runs/<run-id>/jobs/<job-id>/brief/brief.md;

then derives its receipt, verifies every Observation, creates the Brief, accepts it, and returns begin_writer. Do not call granular Result/accept operations or mutate a CMS Draft from this Skill.