Research requirements

Research Requirements

You are tasked with researching requirements for new projects or features. This is the first step in the Research → Plan → Implement workflow, used when starting work that doesn't require understanding an existing codebase.

This command researches and documents:

• Project requirements and constraints
• Technology choices and trade-offs
• Architecture considerations
• External dependencies and existing solutions

Initial Setup

When this command is invoked:

1. Check if parameters were provided:

   • If a tracker URL or project description was provided, proceed immediately
   • Read/fetch input FULLY before spawning sub-tasks

2. If no parameters provided, respond with:

I'm ready to research requirements for a new project or feature. Please provide:
- A project description, OR
- An issue / work-item URL with requirements (any supported tracker)

I'll research technology options, constraints, and existing solutions to inform planning.

Then wait for user input.

Steps to follow after receiving input:

Step 1: Parse Input Source

• If a tracker URL or work-item id was provided:
  • Fetch the work-item content: tracker view <id> --json title,body,state,comments
  • Transition state: tracker set-state <id> research-in-progress
• If plain text: use as-is
• CRITICAL: Read/fetch input FULLY before spawning sub-tasks

Extract:

• Core requirements
• Constraints mentioned
• Technology preferences (if any)
• Success criteria

Step 2: Decompose Research Areas

Break requirements into researchable components:

• Identify functional requirements
• Identify non-functional requirements (performance, security, etc.)
• Identify technology domains needing research

Create a research plan using TodoWrite to track exploration tasks.

Step 3: Spawn Parallel Sub-Agents

Use these agent types concurrently to research different aspects:

Agent	Purpose
web-search-researcher	Technology options, best practices, similar projects
flow-locator	Find existing research in flow/ directory
flow-analyzer	Extract insights from relevant historical docs (if locator finds any)

Run multiple agents in parallel for different research areas. Example:

# Spawn concurrently:
- web-search-researcher: "Research [technology domain 1] options for [use case]"
- web-search-researcher: "Research [technology domain 2] best practices"
- web-search-researcher: "Find similar open source projects for [description]"
- flow-locator: "Find any existing research or decisions about [relevant topics]"

Note on flow agents: Always include these agents. The flow/ directory accumulates historical context over time. For early research cycles, agents may report "no prior research found" - this is expected and still useful information.

Step 4: Wait and Synthesize

• CRITICAL: Wait for ALL sub-agents to complete before proceeding
• Compile findings across all sources
• Identify technology trade-offs and recommendations
• Note open questions requiring user input
• Identify any conflicting recommendations

Step 5: Gather Metadata

Collect metadata for the output document:

• Get current date: date -Iseconds
• Get git commit hash: git rev-parse HEAD
• Get current branch: git branch --show-current
• Get repository name: basename $(git rev-parse --show-toplevel)
• Extract tracker work-item id (if applicable)

Generate filename using this convention:

flow/research/YYYY-MM-DD-gh-[issue-number]-[description].md

• YYYY-MM-DD - today's date
• gh-[issue-number] - tracker work-item id (omit if no work item; the literal "gh-" prefix is a stable filename convention)
• [description] - brief kebab-case description

Examples:

• With issue: flow/research/2026-01-12-gh-1-research-requirements-command.md
• Without issue: flow/research/2026-01-12-dotfile-manager-requirements.md

Step 6: Generate Research Document

Write the research document with this structure:

---
date: [ISO timestamp with timezone]
researcher: claude
git_commit: [current hash]
branch: [current branch]
repository: [repo name]
github_issue: [issue number or null]
topic: "[Project/Feature Name] Requirements Research"
tags: [research, requirements, relevant-tech-tags]
status: complete
---

# Requirements Research: [Project/Feature Name]

## Original Request
[User's input or tracker work-item content]

## Summary
[High-level findings and recommendations - 2-3 paragraphs]

## Functional Requirements
- Requirement 1
- Requirement 2
- ...

## Non-Functional Requirements
- Constraint 1 (e.g., performance, security, scalability)
- Constraint 2
- ...

## Technology Options

### [Domain 1, e.g., "Database"]
| Option | Pros | Cons | Recommendation |
|--------|------|------|----------------|
| Option A | ... | ... | ... |
| Option B | ... | ... | ... |

### [Domain 2, e.g., "Framework"]
| Option | Pros | Cons | Recommendation |
|--------|------|------|----------------|
| Option A | ... | ... | ... |

## Existing Solutions
[Similar projects and approaches from web research]
- [Project/Library Name](url) - Description and relevance
- [Another Resource](url) - Description

## Historical Context (from flow/)
[Relevant past research or decisions, or "No prior research found" for early cycles]

## Open Questions
[Items requiring user clarification before planning]
- Question 1
- Question 2

## Recommended Next Steps
1. Resolve open questions above
2. Run `/create-plan` with this research document

Step 7: Present and Confirm

After writing the document:

1. Transition the work-item state (if applicable):

   • tracker set-state <id> research-complete research-in-progress

2. Present summary to user:

I've completed the requirements research and written the findings to:
`flow/research/[filename].md`

Key findings:
- [Most important finding 1]
- [Most important finding 2]
- [Most important finding 3]

Technology recommendations:
- [Domain]: [Recommended option] because [reason]

Open questions requiring your input:
- [Question 1]
- [Question 2]

Please review the research document. Once you've resolved any open questions, you can proceed with `/create-plan` to design the implementation.

3. Wait for user confirmation before any next steps
4. Do NOT automatically proceed to planning

Important Notes

Critical Ordering

1. Always read/fetch input FULLY before spawning sub-agents
2. Always wait for ALL sub-agents to complete before synthesizing
3. Always gather real metadata (don't use placeholders)
4. Always present findings and wait for user confirmation

What NOT To Do

• DO NOT write any implementation code - this is research only
• DO NOT create a plan - that's /create-plan's job
• DO NOT make final technology decisions - present options with trade-offs
• DO NOT skip user confirmation - always present findings for review
• DO NOT pollute main context - use sub-agents for heavy research
• DO NOT proceed to planning automatically - wait for user approval
• DO NOT use placeholder values - gather real metadata before writing
• DO NOT skip flow agents - always include them, they handle empty results gracefully

References

• Pattern reference: .claude/skills/research-codebase/SKILL.md
• Output directory: flow/research/