devattom/architect

<objective> Analyze codebase architecture, detect structural smells, and safely restructure code while preserving git history and updating all imports. </objective>

<quick_start> Restructure a specific area:

/architect src/components

Full codebase audit + restructure:

/architect -a full codebase restructure

Audit only (no changes):

/architect -s audit auth module

What it does:

1. AUDIT: Scans structure, detects architectural smells (god modules, circular deps, coupling)
2. DESIGN: Proposes restructuring plan with file moves, splits, renames
3. RESTRUCTURE: Executes safely with git mv, import migration, incremental commits
4. VERIFY: Build, tests, import validation </quick_start>

<parameters>

| Flag | Description | |------|-------------| | -a | Auto mode: skip confirmations, execute full workflow | | -e | Economy mode: no subagents, direct tools only | | -s | Save mode: output to .claude/output/architect/ | | --audit-only | Stop after audit step (no restructuring) | | --dry-run | Show proposed changes without executing | | --standalone | Use built-in RESTRUCTURE + VERIFY instead of APEX handoff |

<examples> ```bash /architect src/components # Audit + design → APEX handoff /architect -a reorganize api layer # Auto mode /architect -e simplify auth module # Economy mode /architect -s --audit-only src/ # Audit and save report /architect --dry-run split user module # Preview changes /architect --standalone src/components # Built-in restructure + verify ``` </examples> </parameters> <workflow> ``` DEFAULT: AUDIT → DESIGN → HANDOFF (saves design context, suggests /apex) --standalone: AUDIT → DESIGN → RESTRUCTURE → VERIFY

AUDIT ─ Detect smells, map dependencies, measure complexity DESIGN ─ Propose structure, file moves, module splits HANDOFF ─ Save design context file, offer APEX implementation RESTRUCTURE ─ git mv, update imports, incremental commits (--standalone) VERIFY ─ Build, test, import validation (--standalone)

</workflow>

<state_variables>

| Variable | Type | Description |
|----------|------|-------------|
| `{scope}` | string | Target area to restructure |
| `{task_id}` | string | Kebab-case identifier |
| `{auto_mode}` | boolean | Skip confirmations |
| `{economy_mode}` | boolean | No subagents |
| `{save_mode}` | boolean | Save outputs |
| `{audit_only}` | boolean | Stop after audit |
| `{dry_run}` | boolean | Preview only |
| `{standalone_mode}` | boolean | Use built-in restructure + verify |
| `{smells}` | array | Detected architectural smells |
| `{dependency_map}` | object | File dependency graph |
| `{move_plan}` | array | Planned file operations |

</state_variables>

<reference_files>

| File | When Loaded |
|------|-------------|
| `references/architecture-smells.md` | Always (step-01) |
| `references/architecture-patterns.md` | Step-02 (design phase) |
| `references/restructuring-techniques.md` | Step-03 (execution phase) |

</reference_files>

<entry_point>

Load `steps/step-01-audit.md`

</entry_point>

<step_files>

| Step | File | Purpose |
|------|------|---------|
| 01 | `step-01-audit.md` | Structure analysis + smell detection |
| 02 | `step-02-design.md` | Restructuring plan |
| 03 | `step-03-handoff.md` | Save design context + offer APEX (default) |
| 03-alt | `step-03-restructure.md` | Safe execution (`--standalone` only) |
| 04-alt | `step-04-verify.md` | Build + validation (`--standalone` only) |

</step_files>

<execution_rules>
- Load one step at a time
- NEVER move files without reading them first
- ALWAYS use `git mv` to preserve history
- ALWAYS update imports after every move
- Use parallel agents in step-01 and step-03 (unless economy mode)
- Incremental commits after each logical group of changes
- Build must pass before marking complete
</execution_rules>

<success_criteria>
- Architectural smells correctly identified
- Dependency map accurately captures relationships
- Restructuring plan approved before execution
- All file moves use `git mv` (history preserved)
- All imports updated across codebase
- No circular dependencies introduced
- Build passes without errors
- Tests pass (if tests exist)
- Cleaner, simpler structure achieved
</success_criteria>