auto-review-loop-llm

Automating iterative ML research review loops involves performing a review, applying fixes, and re-reviewing until a positive assessment or maximum rounds is reached. This skill automates that cycle with configurable LLM providers and stateful round tracking.

The autonomous review loop handles state persistence by tracking rounds, scores, and verdicts. It reads prior reviews for context and persists review state to ensure continuous quality improvement across iterative cycles.

Yes, you can use OpenAI-compatible LLM providers with MCP server integration. The loop supports configuration via MCP server or environment variables to integrate with MCP-based LLM chat workflows.

You configure review readiness by setting a POSITIVE_THRESHOLD to determine a positive assessment and MAX_ROUNDS to limit iterations. The loop terminates when either the threshold is met or maximum rounds are reached.

You should not use an autonomous ML review loop when a project lacks explicit scoring and verdict criteria or requires manual subjective judgment. The loop relies on configurable criteria to decide readiness and continuous quality improvement.