Compound Engineering - Brainstorm

What it does

ce-brainstorm structures collaborative conversations to surface what to build. The skill helps teams and product collaborators convert vague ideas or feature requests into a durable requirements artifact (lightweight PRD / feature brief) suitable for handoff to planning. It emphasizes one-question-per-turn dialogue, iterative rigor probes (evidence, specificity, counterfactual, attachment), repository-aware context scanning, and a scoping synthesis checkpoint before writing the final document.

When to use it

Use this skill when a user asks to brainstorm a feature, scope an improvement, or explore product direction - phrases like "let's brainstorm", "what should we build", or when a request is ambiguous or large in scope. It's intended for software/product contexts where the output will guide engineering and planning.

What's included

• Scripts: none executed by the skill itself; the repo includes references and rendering guidelines.
• References: comprehensive references folder guiding rendering, synthesis, and handoff (e.g., markdown/html rendering rules, synthesis-summary, brainstorm-sections).
• Instructions: detailed phased workflow: resume/context scan, product pressure tests, collaborative dialogue with blocking-question tools, approaches + evaluation, scoping synthesis (Path A/B rules), requirements doc generation, vocabulary capture, and handoff options. The skill enforces repo-relative paths in outputs and tier-aware behavior for Lightweight/Standard/Deep scopes.

Compatible agents

Designed for advanced agent frameworks (Claude Code, Codex, Cursor) that can run repo scans and present blocking question UIs; still usable in chat-first agents with careful adherence to its interaction rules.