quick-brainstorm

Lightweight brainstorming for bug fixes, small features, or code improvements that need clarification before implementation. Triggers deep questioning to ensure accurate output. Use when user asks to fix a bug, add a small feature, refactor code, or make targeted improvements — and the scope is small enough that a full brainstorm session is unnecessary. Do NOT use for large-scale architecture decisions or greenfield projects.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "quick-brainstorm" with this command: npx skills add hccake/skills/hccake-skills-quick-brainstorm

Quick Brainstorm

MANDATORY: Do NOT call any file-edit, file-create, or code-generation tools until Phase 1 Q&A self-check has passed or Fast Track criteria are met. Reading files for context is allowed. Violation = immediate stop and return to Q&A.

Lightweight brainstorming → deep Q&A → design confirmation → Plan Mode → execution.

Core principle: Only ask deep, non-obvious questions. Goal is accurate output, not question count.

Optional Parameters

  • --doc — After design confirmation, write design doc to docs/plans/YYYY-MM-DD-<topic>.md using create_file.
  • --worktree — Before starting, create isolated branch via git worktree add ../quick-brainstorm-<topic> -b quick-brainstorm/<topic> and work there.
  • --plan-file — Write implementation plan to docs/plans/YYYY-MM-DD-<topic>-plan.md instead of only presenting in chat.

Process

Task → Understand context (read files/code/commits)
     → Q&A (one question at a time, until self-check passes)
     → Multiple approaches? → Compare options with recommendation
     → Present design in sections, confirm each
     → Enter Plan Mode: create implementation plan, wait for approval
     → Execute after user approval

Fast Track

May skip Q&A only if ALL three conditions are true:

  1. The change is mechanical (single-line fix, typo, rename, or user gave exact code to write)
  2. Zero design decisions or trade-offs involved
  3. User's instruction leaves no room for interpretation

If any condition fails → go to Phase 1. When in doubt, ask one question to verify.

Phase 1: Q&A

Rules:

  • One question at a time, prefer multiple choice
  • Skip anything Claude can infer from code/context
  • Keep asking until self-check passes — not based on a minimum question count

Self-check before ending Q&A:

  • Core functionality approach is clear
  • Edge cases and error handling covered
  • No missing related requirements
  • User has confirmed or implicitly indicated discussion is sufficient (e.g., answered last question with no new concerns)

Phase 2: Compare Options

When: 2+ reasonable approaches exist. Skip when: One clearly optimal solution.

Present each option with pros/cons and a recommendation with reasoning. Keep brief.

Phase 3: Present Design

Present in sections, confirm each before continuing. Trim sections that don't apply.

  1. Change overview — What and why
  2. Technical approach — Implementation details, files involved
  3. Data/API/UI changes — If any
  4. Edge cases — Key cases only
  5. Implementation steps — Brief execution order

Phase 4: Plan Mode

After design confirmation, switch to planning-only mode (no code output yet):

  1. Create implementation plan — List each step: files to change, what to change, expected outcome. If the environment supports a built-in plan mode (e.g. plan command, Plan Mode toggle, or writing-plans skill), use that; otherwise present the plan as a structured list in chat.
  2. Wait for explicit user approval before writing any code.
  3. During execution, pause and ask if encountering scenarios not covered in the plan.

Guardrails

STOP and return to the correct phase if any of these occur:

  • Calling edit/create tools before Q&A self-check passed (unless Fast Track criteria met)
  • User says "no" but continuing with original approach
  • Entering Plan Mode without design confirmation
  • Executing without Plan Mode approval
  • Judging a task as "obvious" without verifying all three Fast Track conditions

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Coding

frontend-design

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.

Repository SourceNeeds Review
94.2K159.5K
anthropics
Coding

remotion-best-practices

Use this skills whenever you are dealing with Remotion code to obtain the domain-specific knowledge.

Repository SourceNeeds Review
2.1K147.4K
remotion-dev
Coding

azure-ai

Service Use When MCP Tools CLI

Repository SourceNeeds Review
155135.8K
microsoft