skills.rest
OpenLAIR
OpenLAIROFFICIAL
Explore all skills by this author →
GitHub

aris-meta-optimize

Analyzes ARIS usage logs to propose data-driven optimizations for SKILL.md prompts and workflow defaults.

1.1k|116|Updated Feb 26, 2026
Search Tags:#self-improvement#prompt optimization#research workflow#llm agents#harness engineering#meta-optimization#usage log analysis
One-click install
npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-meta-optimize
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: aris-meta-optimize
Source: https://github.com/OpenLAIR/dr-claw/tree/main/skills/aris-meta-optimize
Command: npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-meta-optimize

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Over time, an AI research harness accumulates friction: bad default parameters, repeated tool failures, and manual user corrections that signal gaps in skill prompts. This Skill closes that loop by analyzing logged usage events and proposing concrete, evidence-backed patches to the harness itself.

Core Features & Use Cases

  • Usage Pattern Analysis: Computes frequency, failure, convergence, and human-intervention statistics from the .aris/meta/events.jsonl event log.
  • Patch Generation: Produces minimal, one-change-at-a-time diffs for SKILL.md files and workflow defaults, each annotated with the log data justifying it.
  • Cross-Model Review: Sends every proposed patch to an external model (GPT-5.4 via MCP) for adversarial review before recommending it.
  • Use Case: After two weeks of running research workflows, you notice users keep overriding the review score threshold. Run this Skill to confirm the pattern from logs, get a reviewed patch changing the default from 6/10 to 7/10, and apply it with backups.

Quick Start

Ask the assistant to run the meta-optimize analysis on all skills to review accumulated usage logs and propose harness improvements.

Frequently Asked Questions about aris-meta-optimize

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I optimize AI agent skill prompts based on usage data?▼

Enable event logging by merging the provided hooks config into your Claude Code settings, then run the meta-optimize workflow after at least five skill invocations. It analyzes the JSONL event log, ranks optimization opportunities, and generates reviewed patches for your SKILL.md files.

What is outer-loop harness optimization for LLM agents?▼

Outer-loop optimization improves the harness around a model—prompts, defaults, retry rules, workflow ordering—rather than the model weights or the artifacts produced. This approach is inspired by Meta-Harness (Lee et al., 2026), which showed harness design matters as much as model choice.

How much usage data is needed before running meta-optimization?▼

At least five complete skill invocations must be logged in .aris/meta/events.jsonl. The skill checks data availability first and exits with a warning if the log is missing or insufficient, so it never optimizes on noise.

Does the skill apply changes to my workflows automatically?▼

No. Every proposed patch is presented with its supporting evidence and a cross-model reviewer score, and changes are applied only after explicit user approval. Originals are backed up to .aris/meta/backups/ and all changes are logged for reversibility.

What are the limitations of log-driven skill optimization?▼

It cannot optimize artifact schemas or MCP bridge infrastructure config, and it will not propose changes without sufficient log evidence. Patterns observed in only a few runs are flagged as needing more data rather than acted upon.

Related Skills

Explore More Education & Research Skills→
treejhtreejh

session-analyzer

Parses session JSONL logs to verify subagent, hook, and tool behavior against SKILL.md.

Community
Advanced
i-standard1i-standard1
6

skill-auditor

Analyze session transcripts to quantify routing accuracy and cross-skill conflicts.

Official
Advanced
ano4lano4l

loop

Schedule recurring experiment runs with cron and persist loop state.

Community
Advanced
kk20300113-pngkk20300113-png

plan-tune

Configure question prompt frequency and behavioral profiles for gstack skills.

Community
Advanced