Dev Team

Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.

On-demand parallel agent spawning. When a task is too big for one agent and the work can be split into independent streams, spin up multiple engineering agents working in parallel.

Use When

• Multi-file refactor across 10+ files in different modules
• Codebase audit that benefits from parallel exploration
• Performance investigation across multiple subsystems
• Documentation pass across many files

Do Not Use When

• Sequential dependencies (B needs A's output)
• Single-file change
• Task fits within a single agent's context

Workflow

1. Decompose the task into N independent streams
2. Assign each stream to the most appropriate agent (apex / bolt / lens / etc.)
3. Spawn agents in parallel via Task tool
4. Collect results as they return
5. Synthesize into a unified output
6. Save to workspace/development/research/[C]team-{topic}-{date}.md

Streams Example

For "audit the entire auth module":

• Stream 1: @scout-explorer → map all auth files
• Stream 2: @vault-security → security audit
• Stream 3: @lens-reviewer → code quality review
• Stream 4: @grid-tester → test coverage analysis
• Stream 5: @apex-architect → architecture analysis

All spawn in parallel, results combine into one report.

Output

markdown
## Team Investigation — {topic}

### Streams
1. {stream 1 result summary}
2. {stream 2 result summary}
...

### Cross-stream Findings
[Insights that emerged from combining results]

### Recommendation
[Unified next step]

Pairs With

• Any of the 19 engineering agents (you pick which to spawn)
• dev-autopilot (which can spawn its own team internally)
• @compass-planner (often the consumer of team output)