brainstorm_skill

Overview

This skill provides a unified brainstorming entry point with two operational modes: Auto Mode for an end-to-end, multi-role pipeline and Single Role Mode for focused, one-role analysis. It orchestrates interactive framework generation, parallel role analyses, and cross-role synthesis while preserving session continuity and automated task progression. The skill is designed for JSON-driven multi-agent workflows and CLI orchestration.

How this skill works

The skill starts by detecting mode from arguments or asking the user, then routes to Auto Mode or Single Role Mode. Auto Mode runs a 7-step artifacts phase to build a guidance specification, launches parallel role analyses for each selected role, and finishes with an 8-step synthesis that generates feature specs and an index. Single Role Mode runs a 4-step sequential role analysis for one specified role. All phases share session state via workflow-session.json and use a TodoWrite attachment model to auto-continue without manual intervention.

When to use it

• You need a full multi-role brainstorming pipeline from context to feature specs with minimal manual orchestration.
• You want rapid parallel analyses from multiple role perspectives (product, design, architecture, testing).
• You need a focused deep-dive for a single role (e.g., system-architect or ux-expert) within an existing session.
• You want reproducible sessions that persist context and outputs across runs.
• You prefer CLI-driven automation and want to skip interactive prompts using --yes/-y.

Best practices

• Start with Auto Mode for new topics to create a session and full guidance-specification.md before running single-role tasks.
• Use --count to control scope (default 3, max 9) to avoid context overflow and keep synthesis manageable.
• Provide --style-skill for ui-designer tasks to ensure consistent visual direction.
• Check .workflow/active/ for existing sessions to avoid accidental session duplication.
• Use --skip-questions in automated runs but include interactive questions for higher-quality role outputs.

Example use cases

• Run a full pipeline: /brainstorm "Build real-time collaboration platform" --count 3 to generate feature specs and role analyses.
• Quick single-role review: /brainstorm system-architect --session WFS-xyz to produce architecture-specific analysis.md.
• Automated non-interactive run: /brainstorm -y "GOAL: ..." --count 5 to create a session and run all phases end-to-end.
• Design-focused run: /brainstorm ui-designer --session WFS-123 --style-skill material-design to include a visual style skill.
• Recover large contexts by reducing scope: re-run with --count 2 or use --summary-only for synthesis.