Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
cxwudi avatar

Brainstorming

  • 1 installs
  • 1 repo stars
  • Updated August 2, 2026
  • cxwudi/agent-config

Agent Config

About

Agent Config. This repo is the engineering effort of managing any AI Agents related configurations.

  • # Agent Config This repo is the engineering effort of managing any AI Agents re
  • # Agent Config. This repo is the engineering effort of managing any AI Agents re

Brainstorming by the numbers

  • 1 all-time installs (skills.sh)
  • Ranked #2,476 of 3,282 Productivity & Planning skills by installs in the Skillselion catalog
  • Data as of Aug 2, 2026 (Skillselion catalog sync)
npx skills add https://github.com/cxwudi/agent-config --skill brainstorming

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs1
repo stars1
Last updatedAugust 2, 2026
Repositorycxwudi/agent-config

What it does

Agent Config

Files

SKILL.mdMarkdownGitHub ↗

Brainstorming

Help the user think through underdefined or tradeoff-heavy ideas without forcing a durable artifact.

Use this skill when the direction is not settled, when meaningful design trade-offs need to be surfaced, or when the user wants a quick collaborative brainstorm. If the user asks for a written, saved, reviewed, or approved spec, use design-spec instead. If the implementation direction is already clear and no design exploration is needed, use plan instead.

The default output is lightweight: clarified goals, 2-3 meaningful options, a recommendation, a decision summary, or a next-step suggestion. Do not create a spec file unless the user explicitly asks to move into the design-spec workflow.

Boundaries

  • Do not invoke plan, write implementation code, scaffold a project, or take implementation action during the brainstorm.
  • Do not save a durable spec, generate references, dispatch a spec reviewer, or require a user review gate. Those belong to design-spec.
  • If the conversation reveals that the decision is too important or complex to keep informal, recommend switching to design-spec and explain why.
  • Do not transition to plan automatically. Planning is an opt-in next step after the user confirms the direction.

Checklist

You need to follow each of these items and complete them in order:

1. Explore project context: check relevant files, docs, and recent commits as needed for the decision 2. Ask clarifying questions: one at a time, understand goals, constraints, success criteria, and risks 3. Evaluate meaningful design options: present 2-3 viable options with balanced treatment, recommend one, and explain why it is preferred 4. Present and confirm: scale the depth to the complexity, confirm the chosen direction with the user, then offer next steps

Process Flow

flowchart TD
  A["Explore project context"] --> B["Ask clarifying questions"]
  B --> C["Evaluate 2-3 design options"]
  C --> D["Present options and recommendation"]
  D --> E{"User confirms direction?"}
  E -- "no, revise" --> D
  E -- yes --> F{"Offer next steps"}
  F -- "write a spec" --> G(["Hand off to design-spec"])
  F -- "plan low-risk work" --> H(["Hand off to plan"])
  F -- "stop here" --> I(["Done"])

The default terminal state is stopping with a settled direction. Spec writing and planning are opt-in.

Process

Understand the Problem

  • Scale context gathering to the risk of the decision. For codebase work, check relevant files, docs, and recent commits before proposing changes.
  • If the request describes multiple independent subsystems, flag that early and help split the discussion into smaller topics.
  • Ask focused clarifying questions one at a time when goals, constraints, success criteria, or risks are unclear.
  • Prefer multiple choice questions when possible, but use open-ended questions when the user is still shaping the idea.

Explore Options

  • Present 2-3 viable options when meaningful trade-offs exist.
  • Treat options fairly; do not make one option a strawman just to support the recommendation.
  • Recommend one option when appropriate and explain why it fits the user's goals and constraints.
  • If the solution space is genuinely narrow, say so plainly instead of inventing extra options.

Shape the Decision

  • Cover scope, risks, validation considerations, and non-goals at a depth that matches the decision.
  • For quick brainstorming, keep the response compact and useful instead of turning it into a spec.
  • If the user wants to preserve the decision, or the decision is high-risk, cross-cutting, or likely to be resumed later, offer to continue with design-spec.
  • If the user confirms a direction and wants implementation planning for low-risk work that does not need a durable spec, transition to plan.

Design for Isolation and Clarity

  • Break proposed systems into smaller units with clear responsibilities and interfaces.
  • For each unit, clarify what it does, how it is used, and what it depends on.
  • If someone cannot understand what a unit does without reading its internals, the boundaries need work.

Working in Existing Codebases

  • Explore the current structure before proposing changes.
  • Include targeted improvements only when existing problems affect the current goal.
  • Do not propose unrelated refactoring.

Next Steps

Once the user confirms the direction, summarize the settled design briefly and offer the next step. Default to stopping; do not run anything automatically.

"Design settled. Want me to write a durable spec (design-spec), go straight to a plan for this low-risk work (plan), or stop here?"
  • If the user wants a saved, reviewed spec, hand off to design-spec.
  • If the user wants implementation planning and the work does not need a durable spec, hand off to plan.
  • If the work is high-risk, cross-cutting, or likely to need review, recommend design-spec before plan.
  • Otherwise, stop. The settled direction lives in the conversation.

Key Principles

  • Keep it lightweight: the default output is a conversation, not a file
  • Ask one question at a time.
  • Present options fairly and recommend clearly.
  • Offer, do not force: artifacts and planning are opt-in next steps.
  • State assumptions when moving quickly.
  • Stay focused on the current decision.

Diagrams

When visual explanation would help, use lightweight Mermaid diagrams directly in the conversation. Mermaid works well for architecture diagrams, flowcharts, and sequence diagrams. If a diagram is awkward to express in Mermaid, use an ASCII diagram instead.

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.