/issue — GitHub Issue Analysis & Planning
Analyse a GitHub issue, explore the codebase, clarify ambiguities, and produce a detailed implementation plan. Stops before implementation — hand off to /dev when ready.
Phases
| # | Phase | What happens |
|---|---|---|
| 1 | Fetch | Parse input (number or URL), fetch issue via gh issue view |
| 2 | Detect | Identify tech stack from manifest files |
| 3 | Explore | Delegate to the built-in Explore subagent, get affected files and patterns |
| 4 | Clarify | Present understanding, ask questions — STOP for user input |
| 5 | Plan | Create implementation plan with AI risk assessment |
| 6 | Todo | Generate task list via TodoWrite, then STOP |
Key Constraints
- MUST fetch the issue using
ghCLI - Delegate wide codebase exploration to the built-in Explore subagent; explore directly only when the area is already known
- MUST stop after creating the todo list — do NOT implement
- Use sequential-thinking MCP for complex planning decisions
- Use context7 MCP for library documentation lookups
Handoff
When complete:
/dev "Implement #[issue-number]" @[plan file if saved]
Full Workflow
For complete phase-by-phase instructions, gh commands, exploration prompts, plan format, and AI assessment template, see references/workflow.md.