meta-optimize

Optimize research harness performance by analyzing accumulated ARIS execution logs to identify workflow bottlenecks, redundant model-compensation steps, and inefficient defaults. The analysis targets structural improvements for skill prompts and harness scaffolding to resolve pipeline stagnation issues.

Harness dieting is the process of proposing the removal of outdated model-compensation scaffolding from research workflows. You need it when newer models render specific compensation steps unnecessary, allowing the harness to operate more efficiently without redundant instructions.

Identify ARIS research pipeline bottlenecks by synthesizing event log data to pinpoint the single most limiting stage, such as planning or verification quality. The analysis evaluates usage patterns, high-frequency skill invocations, and common failure points to isolate the constraint.

Analyzing ARIS usage logs requires access to the .aris/meta/events.jsonl file to process execution data with Bash and Python. If the events.jsonl file is missing or empty, the log analysis cannot be performed and no optimization report can be generated.

The best way to analyze skill invocations and user intervention patterns is to evaluate accumulated event logs using Python for data synthesis and Bash for log processing. This usage pattern analysis reveals common failure points and highlights potential structural optimizations for prompts.

Research workflow stagnation happens when inefficient defaults and redundant scaffolding accumulate over time. Log analysis resolves it by evaluating event logs to identify performance bottlenecks and propose structural optimizations for skill prompts and harness scaffolding.