CodeLeash: framework for quality agent development, NOT an orchestrator
TDD state machine leash for Claude Code avoids agent drift, but niche audience.
Quality gates for AI agents. Guards that don't get tired.
465 passing tests from 5,000+ production executions—guards that don't get tired.
AI agent developers, ML engineering teams, automation engineers
Guardrails AI · Llama Guard · LangChain Guards
TDD state machine leash for Claude Code avoids agent drift, but niche audience.
tmux-native agent orchestration with git worktrees beats CrewAI's invisible background processes.
AST-verified AI code audits prevent hallucinations; LLM findings checked against parser ground truth.
17-agent team with session memory, but Anthropic Batch and Claude Projects already persist context.
Replaces flaky LLM judges with strict Python equality checks for tool arguments.
Flips AI coding on its head: human gates decide, not the AI's confidence.