claude-code-fsm-workflow0 Multi-agent workflow for Claude Code where discipline is enforced by hooks, not personas. 23 subagents with strict role separation, context isolation, and nonce-proof reads — plus deterministic AST audit scripts that replace LLM auditors. Agents can't drift, contradict the spec, or claim work done without proof.
1 plugin · itsaldrincr/claude-code-fsm-workflow