hellblazer/codebase-analysis

Codebase Analysis Skill

Delegates to the codebase-deep-analyzer agent (model: sonnet).

When This Skill Activates

• Exploring a new codebase or module before making changes
• Onboarding to a project — understanding structure and patterns
• Trying to understand how a feature works or where something is defined
• Before undertaking major refactoring
• Mapping dependencies between components or modules

Pre-Dispatch: Seed Link Context (optional)

If the analysis is for a specific RDR or architectural decision, seed link-context so findings stored to T3 auto-link. See /nx:catalog for details. Skip for general exploration.

Agent Invocation

Use the Agent tool to invoke codebase-deep-analyzer:

## Relay: codebase-deep-analyzer

**Task**: [what needs to be done]
**Bead**: [ID] or 'none'

### Input Artifacts
- Files: [relevant files]

### Deliverable
Architecture analysis with patterns and dependencies

### Quality Criteria
- [ ] Architecture overview documented
- [ ] Key patterns identified and explained
- [ ] Dependency relationships mapped

For full relay structure and optional fields, see RELAY_TEMPLATE.md.

Analysis Methodology

The codebase-deep-analyzer uses mcp__plugin_nx_sequential-thinking__sequentialthinking:

1. Form hypothesis about architecture (e.g., "This appears to be MVC pattern")
2. Gather evidence from code structure, naming, dependencies
3. Validate/refute hypothesis against actual code
4. Map module/package structure and document patterns
5. Persist findings to nx store: mcp__plugin_nx_nexus__store_put(content="...", collection="knowledge", title="architecture-{project}-{component}", tags="architecture"

Success Criteria

• [ ] Architecture overview documented
• [ ] Module/component map created
• [ ] Key patterns identified and explained
• [ ] Dependency relationships mapped
• [ ] Findings stored in nx store for future reference
• [ ] Subtask findings promoted from T1 to T2

Agent-Specific PRODUCE

Outputs generated by the codebase-deep-analyzer agent:

• T3 knowledge: Architecture findings via store_put tool: content="# Architecture: {project}/{component}\n{findings}", collection="knowledge", title="architecture-{project}-{component}", tags="architecture,analysis"
• T2 memory: Subtask findings promoted via scratch_manage tool: action="promote", entry_id="<id>", project="{project}", title="analysis-{component}.md"
• T1 scratch: Per-subtask findings via scratch tool: action="put", content="Subtask {N} findings: {summary}", tags="analysis,subtask-{N}" (flagged for T2 at phase end)

Session Scratch (T1): Agent uses scratch tool for ephemeral working notes during the session. Flagged items auto-promote to T2 at session end.