Spec-driven development workflow: scout codebase, research docs, plan tasks, and execute
Debug and analyze LangChain agents using LangSmith traces via CLI export and Python analysis scripts
Orchestrate external CLI coding agents (Qwen Code, Gemini CLI, OpenCode CLI) for parallel execution, diverse perspectives, and distributed LLM quota usage
Two-agent adversarial code review with a Sonnet critiquer that debates design decisions, finds logical flaws, and negotiates net actions before any code changes