Claude Code Gets a Hub-and-Spoke Collective of TDD Agents

What it does

claude-code-collective is an NPX package that scaffolds 30+ specialized agent definitions, shell hooks, and a behavioral rulebook into your project. It turns Claude Code into a test-first development team with a central router (/van → @task-orchestrator) that delegates tasks to specialists. The collective enforces a RED → GREEN → REFACTOR cycle and uses Context7 to pull real library documentation instead of guessing.

The interesting bit

The project treats Claude Code less like a chatbot and more like an operating system that needs guardrails. The author built a “behavioral operating system” (CLAUDE.md) and shell hooks that intervene before code generation to enforce test-first development, addressing common agent failures like mid-task abandonment and API hallucination with infrastructure rather than pleading.

Key highlights

• Hub-and-spoke routing: one orchestrator delegates to 30+ specialist agents instead of letting them self-select.
• Context7 integration pulls real library documentation rather than letting agents guess.
• TDD enforcement via shell hooks and handoff contracts that block completion until standards are met.
• Includes quality gates, a research analyzer for complexity scoring, and Playwright-based functional testing.
• Honest about being an experimental personal MVP tool, not enterprise software.

Caveats

• Requires a Claude Code restart after installation for hooks to load, and the research phase can be slow.
• Documentation is scattered across files, and some agents are still being refined.
• Deeply opinionated about test-first development; if you skip tests, this will fight you.

Verdict

Worth a look if you use Claude Code daily and wish it had the discipline to write tests before implementation. Skip it if you prefer lightweight, unopinionated assistants or need production-grade stability.