From superpowers
Explores project codebase and context extensively, identifies requirement gaps, then asks one focused clarifying question at a time to build a clear mental model.
How this command is triggered — by the user, by Claude, or both
Slash command
/superpowers:discoverybrainstorming/references/The summary Claude sees in its command listing — used to decide when to auto-load this command
# Phase 1: Discovery - Detailed Guidance ## Goal Understand what's being built by exploring the current project state and clarifying requirements. ## Primary Agent Actions ### 1. Explore Codebase First **Before asking any questions**, build context from existing code: - Find relevant files matching patterns (e.g., `**/*.ts`, `src/**/*.py`) - Search for patterns and similar implementations - Read files to understand existing code structure and conventions - Check for architectural patterns, naming conventions, error handling styles ### 2. Review Project Context Understand the project...
Understand what's being built by exploring the current project state and clarifying requirements.
Before asking any questions, build context from existing code:
**/*.ts, src/**/*.py)Understand the project's established patterns:
docs/ directory for existing documentationREADME.md for project overview and setupCLAUDE.md for development guidelines and constraintsgit log --oneline -20 to see recent commits and development focusBased on your exploration, identify what's unclear:
Use AskUserQuestion tool strategically:
CRITICAL: One Question at a Time:
AskUserQuestion with exactly 1 question (never use 2-4 questions)Question Style Preferences:
Question Categories:
Example Multiple Choice Questions:
I found two existing authentication patterns in the codebase:
1. JWT tokens with 24-hour expiry (used in /api/v1)
2. Session-based auth with Redis (used in /admin)
Which approach should this feature use?
A) JWT tokens (stateless, scales better)
B) Session-based (more secure, easier revocation)
C) Different approach (please describe)
Synthesize exploration and user answers:
Explore extensively before asking. Questions should fill gaps that codebase exploration cannot answer. If you can learn it from code or docs, don't ask.
Clear understanding of requirements including:
To clarify scope:
"I see the current user model has
phone. Should the new notification system support SMS (usingphone) or just Email?" (Good because it links to specific existing fields)
To clarify edge cases:
"The current
calculateTotalfunction throws an error for negative inputs. Should this new API return a 400 Bad Request in that case, or clamp the value to zero?"
The "Lazy" Question:
"How should I implement this?" Why it's bad: You should explore the codebase and propose options, not ask for implementation details.
The "Bundled" Question:
"Should I use Redis? Also what about the schema? And do we need a migration?" Why it's bad: Overwhelms the user. Ask about the storage choice first.
The "Abstract" Question:
"What are the non-functional requirements?" Why it's bad: Too vague. Ask specific questions like "Do we need to support more than 1000 requests/second based on the current load balancer config?"
npx claudepluginhub fradser/dotclaude --plugin superpowers/requirements-startStarts multi-phase requirements gathering for a feature slug from $ARGUMENTS: analyzes codebase, asks discovery questions, investigates context autonomously, and generates expert queries, producing structured docs.
/sc-workflowGuides feature development via phases: codebase exploration with agents, clarifying questions, architecture design, and step-by-step implementation.
/discussConducts thorough technical discussion on a topic: scans codebase context, asks clarifying questions in batches, summarizes requirements, and recommends next steps.
/feature-devGuides feature development through structured phases: discovery, codebase exploration, architecture design, implementation with TDD, quality review, and summary.
/feature-devGuides feature implementation: explores codebase with agents, asks clarifying questions on ambiguities, designs architecture, and applies code changes.
/kickoffConducts deep codebase research, external best practices review, and MVP pathfinding, producing a kickoff document with architecture insights, gap analysis, and implementation options.