Research Report Pipeline

End-to-end multi-agent research → draft → review pipeline that exercises every multi-agent authoring primitive. Demonstrates the canonical pattern tenant ecosystems can fork for branded authoring workflows. Phases: 1. Plan — single agent emits an outline + open questions to scratchpad 2. Research — parallel_subagent_fanout spawns N Bot subagents on the outline's open_questions; results land in scratchpad slot `research_findings` 3...

Details

End-to-end multi-agent research → draft → review pipeline that exercises every multi-agent authoring primitive. Demonstrates the canonical pattern tenant ecosystems can fork for branded authoring workflows. Phases: 1. Plan — single agent emits an outline + open questions to scratchpad 2. Research — parallel_subagent_fanout spawns N Bot subagents on the outline's open_questions; results land in scratchpad slot research_findings 3. Draft — parallel section drafts (one per outline section) all writing to scratchpad slot draft_sections 4. Review — single agent reviews assembled draft; verdict + issues persisted to scratchpad decisions 5. Publish — assembles draft sections into a Content row with authorship attribution This workflow is referenced by the research-report Skill, so callers can either invoke it directly OR call execute_skill(slug='research-report', ...) and get the same multi-step execution with proper SkillExecution accounting.