meta-iterate

What problem does it solve?

Iterative improvement of Agents and Skills requires coordinating tests, prompt changes, reviews, backups, and state recovery across multiple rounds, which is time-consuming and error-prone when done manually. meta-iterate automates the orchestration of test → optimize → review → re-test loops, preserves reproducible plan files, and enforces backup and synchronization to ensure safe, resumable optimization.

Core Features & Use Cases

• Multi-phase strategy: supports warmup, baseline, sampling, and full verification strategies depending on testcase scale.
• End-to-end orchestration: generates timestamped iterate plans, recovers interrupted runs, invokes meta-prompt-engineer and meta-reviewer subagents, and writes learnings and changelogs.
• Operational guardrails: enforces backups, plan persistence, status sync, and strict stop/rollback rules to avoid regressions.
• Use Case: accelerate improvement of a translation Agent by running warmup samples, creating a baseline, iterating on sampled cases, and validating changes on the full testcase set before applying them.

Quick Start

Start an iterative optimization run for the agent named my-agent in source/agents/my-agent on platform codebuddycli with target_score 95.