Run iterative improvement loops with JSON evaluation and decision logs.
npx skills add https://github.com/xz1220/oh-my-kimi --skill autoresearch-xz1220
Please help me install this Agent Skill. Skill: autoresearch Source: https://github.com/xz1220/oh-my-kimi/tree/main/skills/autoresearch Command: npx skills add https://github.com/xz1220/oh-my-kimi --skill autoresearch-xz1220
It solves the problem of improving a single mission through repeated experimentation while keeping evaluation results consistent, auditable, and bounded by an explicit stopping rule.
Run autoresearch after you obtain a mission and evaluator from deep-interview, and let it persist iterations under .omk/autoresearch/ while enforcing the max-runtime bound.
High-intent search queries and answers about installing and using this skill.
Iterative improvement requires a strict evaluator contract where each run outputs structured JSON with a boolean pass field and optional score, while appending a human-readable decision log to track why changes occurred. This ensures evaluations remain consistent and auditable across repeated experiment cycles.
Persisting experiment tracking data involves saving machine-readable evaluation JSON and human-readable markdown decision logs under a dedicated directory. This preserves run history for stateful agents without overwriting earlier evidence during subsequent periodic reruns.
To stop an iterative experimentation loop from running indefinitely, enforce an explicit max-runtime bound or another explicit termination condition. The loop halts only when this threshold is reached rather than silently running forever.
Yes, you can schedule periodic reruns for a stateful single-mission improvement loop using cron. This compatibility allows recurring executions without overwriting earlier machine-readable JSON evaluation evidence or markdown decision logs.
JSON evaluation for iterative experimentation requires a structured output format containing a boolean pass field and an optional numeric score. This strict evaluator contract ensures each iteration is recorded consistently as machine-readable evidence.
xdkpIterate on a mission using evaluator-driven pass/fail JSON until max-runtime ends.
jimmi2051Run evaluator-driven mission improvement loops with JSON logs and Markdown decision records.
profzellerScaffold RunPod Serverless projects with handler.py, Dockerfile, and hub.json.
MohammedSaudAlsahliEvaluate, record, and repeat single missions with structured JSON outputs and markdown logs.
