brainstorming

Refines ideas into approved designs through structured one-question-at-a-time dialogue before implementation.

Updated Aug 2, 2026
One-click install
npx skills add https://github.com/leonardoacosta/skills --skill brainstorming-leonardoacosta
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: brainstorming
Source: https://github.com/leonardoacosta/skills/tree/main/leo-core/skills/brainstorming
Command: npx skills add https://github.com/leonardoacosta/skills --skill brainstorming-leonardoacosta

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Teams and agents often jump straight into coding on unexamined assumptions, producing wasted work and misaligned features. This Skill enforces a design-first gate: it explores project context, asks clarifying questions one at a time, proposes 2-3 approaches with trade-offs, and requires explicit user approval of a written design before any implementation begins. ## Core Features & Use Cases - Structured requirements dialogue: Explores existing files, docs, and commits first, then asks one question per message with a recommended answer, covering purpose, constraints, and success criteria. - Design approval gate: Presents the design section by section (architecture, components, data flow, error handling, testing), persists it to OpenSpec changes or a single Beads issue, runs a spec self-review, and blocks implementation until the user approves. - Visual companion server: Ships a Node.js WebSocket server (scripts/server.cjs) that serves HTML mockups, wireframes, and A/B option screens to a browser so visual questions can be answered by clicking instead of reading. - Use Case: Before building a new dashboard feature, the agent checks the repo, asks whether real-time updates are required, proposes polling versus WebSocket approaches, presents the design in sections, and only hands off to the implementation workflow after you approve the written spec. ## Quick Start Ask the agent to brainstorm and design the feature you have in mind before writing any code, and answer its clarifying questions one at a time until it presents a design for your approval.

Frequently Asked Questions about brainstorming

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I brainstorm a feature design with an AI agent before coding?

Invoke the brainstorming flow before any implementation: the agent explores your project context, asks clarifying questions one at a time, proposes 2-3 approaches with trade-offs, and presents a design in sections. No code is written until you approve the written design.

What is the best way to gather software requirements through dialogue?

Ask one question per message, prefer multiple-choice formats, and lead each question with a recommended answer and reasoning. Look up discoverable facts in the codebase yourself and reserve questions for genuine decisions only the user can make.

Can I show UI mockups in a browser during brainstorming?

Yes. The visual companion runs a local Node.js server (scripts/start-server.sh) that serves HTML mockups, wireframes, and A/B option screens to a browser. User clicks are recorded as JSON events that the agent reads on the next turn.

Does the brainstorming skill work with very small tasks like a config change?

Yes, the design gate applies to every task regardless of size. For simple tasks the design can be a few sentences, but it must still be presented and approved before implementation to surface unexamined assumptions.

What happens after the design is approved in brainstorming?

The approved design is persisted to an OpenSpec change directory or a single Beads issue, goes through a spec self-review for placeholders and contradictions, and is then handed to the repository's execution workflow such as feature or apply.

When should I use the design-it-twice parallel protocol?

Use it at the design gate only when the interface shape itself is contested and expensive to reverse, such as a public API or package boundary. It spawns 3+ parallel agents each pinned to a different constraint, then compares their designs with one opinionated recommendation.