research-phase

---

name: research-phase description: Research how to implement a phase before planning argument-hint: "[phase]" allowed-tools:

• Read
• Bash
• Glob
• Grep
• Write
• WebFetch
• WebSearch
• mcp__context7__*

---

This is for niche/complex domains where Claude's training data is sparse or outdated. Research discovers:

• What libraries exist for this problem
• What architecture patterns experts use
• What the standard stack looks like
• What problems people commonly hit
• What NOT to hand-roll (use existing solutions)

Output: RESEARCH.md with ecosystem knowledge that informs quality planning.

<execution_context> @/.claude/ag4one/references/principles.md @/.claude/ag4one/workflows/research-phase.md @/.claude/ag4one/templates/research.md @/.claude/ag4one/references/research-pitfalls.md </execution_context>

Load project state: @.planning/STATE.md

Load roadmap: @.planning/ROADMAP.md

Load requirements: @.planning/REQUIREMENTS.md

Extract phase requirements before research:

1. Find the phase in ROADMAP.md, get its Requirements: list
2. Look up each REQ-ID in REQUIREMENTS.md for full description
3. Use concrete requirements to focus research domains

Load phase context if exists: Check for .planning/phases/XX-name/{phase}-CONTEXT.md - bonus context from discuss-phase.

<success_criteria>

• Phase validated against roadmap
• Domain/ecosystem identified from phase description
• Comprehensive research executed (Context7 + official docs + WebSearch)
• All WebSearch findings cross-verified with authoritative sources
• RESEARCH.md created with ecosystem knowledge
• Standard stack/libraries identified
• Architecture patterns documented
• Common pitfalls catalogued
• What NOT to hand-roll is clear
• User knows next steps (plan phase) </success_criteria>