analyze

@analyze

Trigger: Before a large refactor, investigating performance, finding tech debt, or understanding a complex area.

Protocol

1. Read only — no code changes.

2. Identify the scope from the argument. For brownfield: scan the relevant files first.

3. Analyze each dimension:

   Complexity hotspots

   • Functions > 40 lines
   • Nesting depth > 4
   • Files > 300 lines
   • Long parameter lists (> 4 args)

   Coupling

   • What imports what — who depends on this module?
   • Blast radius: if this file changes, what breaks?
   • Circular dependencies

   Dead code

   • Unused exports
   • Unreachable branches (if (false), code after return)
   • TODO/FIXME/HACK comments — count and locate

   Test gaps

   • Which functions/branches have no test coverage?
   • Which error paths are untested?

   Dependencies

   • Outdated packages (compare versions)
   • Packages imported but barely used

4. Report:

   ## Analysis: {scope}
   
   ### Overview
   [2-3 sentences]
   
   ### Hotspots (top 5 by severity)
   1. `file:line` — issue, why it matters, recommended action
   
   ### Dependency map
   [key relationships and blast radius]
   
   ### Tech debt inventory
   TODOs: N | Complexity flags: N | Dead code: N
   
   ### Recommended actions
   Priority 1 (do now): ...
   Priority 2 (do soon): ...
   Priority 3 (backlog): ...
   

5. Call save_artifact type=specs, slug=analyze-{scope-slug}.

6. Call set_memory key=analyze-{scope}, value=the report.

Rules

• Read-only. No code changes.
• Rank findings by impact, not volume.
• "Recommended actions" must be concrete and actionable, not vague ("consider refactoring").