meta-optimize

Usage log analysis optimizes SKILL.md files by parsing `.aris/meta/events.jsonl` for failure rates and user interventions, then proposing targeted diff patches to improve workflow defaults. It requires at least 5 complete workflow runs to generate actionable data.

Generating code diffs for ARIS configurations involves analyzing event logs with Python and using GPT-5.5 for adversarial review. This cross-model review ensures proposed patches have minimal impact on existing workflows before they are applied.

Yes, analyzing ARIS usage logs requires a minimum of 5 complete workflow runs logged in `.aris/meta/events.jsonl`. This threshold ensures sufficient data exists to identify meaningful usage patterns, failure rates, and user interventions.

Yes, you can use Python for data analysis and Bash for script execution to automate reviewer prompt optimization. The process evaluates logged events to propose data-driven changes to prompts and convergence rules.

GPT-5.5 provides adversarial review of generated diff patches to ensure minimal impact on existing workflows. This cross-model evaluation validates that proposed changes to SKILL.md files and default parameters maintain system stability.

Optimization proposals fail when `.aris/meta/events.jsonl` lacks sufficient data, specifically fewer than 5 complete workflow runs. Without adequate logs, the analysis cannot detect meaningful failure rates or user interventions to propose changes.