Self-orchestrating, parallel-first multi-agent development system for Claude Code — orchestrated by best / Claude Opus 4.8 at ultracode effort. 15 agents (8 core + 1 security gate + 6 department), Smart Routing default, dynamic-workflows escalation, risk-based quality gates. You say WHAT, the AI decides HOW.
Optional department agent for GitHub Actions, repository protection, and CI/security guardrails. Invoke when touching .github/workflows/, CODEOWNERS, dependabot.yml, or any repository security surface.
Read-only documentation and developer-experience reviewer. Invoke when public-facing docs, prompts, setup instructions, or user-facing clarity are in scope.
Read-only quality-operations specialist for validation scope, regression gates, and eval-oriented workflow checks. Invoke when defining the test plan or validation strategy for a changed scope.
Read-only runtime and platform specialist for toolchain, sandbox, environment, and OS behavior. Invoke when environment constraints, toolchain prerequisites, or OS-specific behavior affect a task.
Read-only workflow designer for orchestration, skill boundaries, prompt structure, and durable handoff artifacts. Invoke when the task involves changing how agents coordinate, skill routing, or handoff reliability.
Experimental Agent Teams orchestration — run CCGodMode agents as parallel teammates with SharedTaskList coordination (requires CLAUDECODEEXPERIMENTALAGENTTEAMS=1).
Ultracode / Max-Parallel mode — dynamic workflows fan work out across tens–hundreds of adversarially-verified parallel subagents for large, decomposable jobs (codebase-wide audits, big migrations, cross-checked research). Opt-in; higher token spend.
Bootstrap repo-local governance before CCGodMode workflows run in an empty, newly initialized, or undocumented workspace. Use proactively when a project has no CLAUDE.md, no README, or no clear structure yet.