from ideation_team_skill
Orchestrates multi-agent ideation sessions (Free Thinker, Grounder, Writer, optional Explorer) to explore concepts, capture source materials, and produce struct
This skill runs structured multi-agent ideation sessions: it interviews the user to scope a concept, captures all source materials, spawns specialized agents (Free Thinker, Grounder, Writer, optional Explorer) as an Agent Team, coordinates depth-aware dialogue, and produces deliverables (vision document, idea briefs, distribution page, presentation, PDFs). It covers session lifecycle: Plan (configure + interview), Ideate (orchestrate team, evaluate idea reports, trigger production), Continue (versioned resume), and PRD (generate a product requirements document from completed session output).
Use this skill when you need fast, structured creative exploration of a concept that must be documented and packaged for distribution or handoff. Good for product discovery, early-stage concepting, internal strategy workshops, or when you want reproducible, versioned ideation outputs. It is designed for Agent Teams (Claude Code) workflows and for scenarios that benefit from separate generative, convergent, and synthesis roles.
Primarily designed for Claude Code Agent Teams (agent-teams), but the patterns and spawn prompts can be adapted to other multi-agent runtimes that support task lists, message broadcasting, and file output (e.g., Opus-like systems).
Multi-agent ideation skill for Claude Code Agent Teams. Orchestrates Free Thinker, Grounder, Writer, and Explorer agents through Plan/Ideate/Continue/PRD actions. Exceptionally well-documented with detailed agent spawn prompts, depth directives, and production phase workflows. No scripts included — purely instructional SKILL.md (74K chars). Requires experimental Agent Teams feature, limiting practical usability.
Purely instructional skill with no executable scripts. No security concerns whatsoever. The main issues are: (1) extreme length makes it hard to consume, (2) requires an experimental Claude Code feature that most users won't have enabled, (3) monolithic structure could benefit from splitting into references/ and templates/. The quality of writing and depth of instruction is excellent — the skill is very thorough in guiding the Arbiter agent through each phase.
