coco

CoCo Super Intelligence is an open-core orchestration layer that installs into existing AI coding tools — Claude Code, Cursor, Codex CLI, and any AGENTS.md-compatible editor — upgrading them with an expert advisory board, a skills library, persistent project memory, specialized subagents, and verification gates. The core offering is Super Intelligence, a routed advisory board of 495 named expert personas across 13 departments (Engineering, AI, Product & Design, Finance, Trading, Legal & IP, Risk & Compliance, Strategy, Data & Analytics, Climate & Energy, Education, Healthcare, and Sales/GTM/Marketing). Commands like `/SI-Decide`, `/SI-AI-Decide`, and `/SI-Eng-Pre-Mortem` invoke panels that deliberate in reacting rounds before returning attributed verdicts with cited public sources. A two-stage routing algorithm selects relevant teams and persons without loading the full filesystem at runtime. Three architectural pillars distinguish CoCo: 1. **Autonomous team execution** — `/team:ship` runs a coordinated build pipeline with six development stages, seven hard verification gates, and two architecture conformance gates. Specialized agents (Research, Architecture, Plan, Review, Build, QA) act in their domains. Code must pass the Test Evidence Protocol and land where planned before a pull request opens. 2. **Disk-persistent state** — Phase state, decisions, and progress logs survive `/clear` resets via atomic git commits, enabling rollback or resumption from any checkpoint. 3. **Vendor-neutral portability** — Rules, templates, and agent definitions are pure Markdown and YAML; IDE adapters inject the same library into Claude Code, Cursor, Codex CLI, VS Code, and similar tools. Architecture conformance checking closes a gap that tests alone cannot: it detects when components are relocated, folded, or deleted without breaking behavior. A commit-pinned index at `.arch/index.json` maps every component to its real filesystem paths. A `built-as-declared` gate verifies that the final build matches the declared architecture plan, blocking silent deviations. The install includes 227 skills (75 core + 152 bundle), 386 slash commands, 35 specialized agents, and 15 cross-IDE rules. A plain install ships 145 active assets; the security-testing `reverse-skill` bundle is opt-in. System bundles extend the workspace: - **GSD (Get Shit Done):** 68 skills and 24 agents managing complex codebase lifecycles — disk-backed phase databases, isolated git workstreams, post-mortem forensics, and autonomous multi-wave execution. - **Brain:** A local knowledge-graph engine connecting email, chat, code, and documentation via SQLite, with a wiki generator and mail sync. - **Team:** Multi-agent product pipeline (`/team:ship`, `/team:plan`, `/team:review`, `/team:verify`) with deterministic merge gates running lints, import audits, and verification suites. - **Super Intelligence:** The 495-persona advisory board and its 342 generated commands. Installation runs in approximately 90 seconds via `npx cocosuperintelligence`, cloning the pinned release locally with no telemetry. The core is MIT-licensed; Super Intelligence is proprietary for own-use.