Claude Code setup with specialist agents, enforcement hooks, and MCP-powered context. How I build the other projects.
A Claude Code configuration with 6 specialist sub-agents (adversarial, auditor, guardian, supervisor, team-lead, infra-fixer), 10+ skills, and automated hooks that enforce architectural rules. The AL MCP server gives agents access to symbol data, event graphs, and code intelligence without burning context on file reads. Implementation sessions run largely autonomously.
How I use Claude Code's agents, skills, rules, and hookify to enforce architecture boundaries and run autonomous TDD loops on a 12-crate Rust workspace.
21 commands backed by an in-memory symbol index. Event tracing, call graphs, impact analysis, and an agentic debugger that Microsoft's MCP doesn't cover.