oh-my-codex: Skill for Codex

Ultrawork Skill (ulw)

Parallel execution mode for large-scale multi-file operations.

When to Use

• Refactoring across 5+ files
• Bulk updates (imports, naming conventions)
• Parallel feature development (independent components)
• Large codebase migrations

When NOT to Use

• Sequential dependencies between changes
• Single-file modifications
• Tasks requiring careful ordering
• When changes might conflict

Workflow

1. DECOMPOSE
   - Identify independent work units
   - Group by file/module boundaries
   - Check for conflicts between units

2. PARALLELIZE
   - Spawn sub-tasks (conceptually parallel)
   - Each unit gets isolated context
   - No cross-unit dependencies allowed

3. AGGREGATE
   - Collect all results
   - Check for merge conflicts
   - Resolve any inconsistencies

4. VERIFY
   - Run full test suite
   - Check for broken imports
   - Validate integration points

Decomposition Rules

INDEPENDENT if:
- Different files with no shared imports
- Same file, different functions (careful!)
- Different modules entirely

DEPENDENT if:
- Shared type definitions
- Import/export relationships
- Shared state or config

Task Unit Template

## Unit: {name}
Files: {file1}, {file2}
Action: {what to do}
Isolated: {yes/no}
Conflicts with: {none | unit names}

Execution Strategy

For truly parallel work with Codex MCP:

# Conceptual - actual implementation in orchestrator/
async def ultrawork(units):
    tasks = [spawn_codex_session(unit) for unit in units]
    results = await asyncio.gather(*tasks)
    return merge_results(results)

Examples

Good trigger:

"ulw: rename all instances of 'userId' to 'user_id' across the codebase"

Good trigger:

"ulw: add TypeScript types to all utility functions"

Bad trigger:

"ulw: refactor the auth module" (likely has dependencies)