harness-evolver
Reduces manual tuning of Claude agent configurations by automatically diagnosing failure modes and proposing optimized prompts, routing rules, and retrieval …
Automated harness evolution for AI agents. A Claude Code plugin that iteratively optimizes system prompts, routing, retrieval, and orchestration code using full-trace counterfactual diagnosis. Based on Meta-Harness (Lee et al., 2026).
- Auto-tune routing logic when agent picks wrong tool for task
- Optimize retrieval augmentation when RAG returns irrelevant context
- Refine system prompt based on trace analysis of failure patterns
Reduces manual tuning of Claude agent configurations by automatically diagnosing failure modes and proposing optimized prompts, routing rules, and retrieval strategies. Cuts iteration cycles from weeks to hours.
Teams deploying multi-step Claude agents who need faster optimization of system prompts and orchestration logic without hand-tuning each component.
https://github.com/raphaelchristi/harness-evolver
By raphaelchristi
How to Get It
claude plugins install raphaelchristi/harness-evolver
Tip: Paste this into a Claude Code conversation. Verify command matches your Claude Code version.
Auto-generated from the tool's public listing — not hands-on verified. Cross-check against the source repo's README before running.
After installing, paste this into Claude:
Help me auto-tune routing logic when agent picks wrong tool for task
Trust Signals Auto-scanned
Community Pulse Growing
Discussed on Reddit
- These fabricated creatures make surreal dreamscapes as they combine engineering — Reddit · 2021 pts
- Over 700 miles miles and things are looking good, RS3 transmission is happy, OEM — Reddit · 798 pts
- Chernobyl fungus appears to have evolved the ability to harness ionizing radiati — Reddit · 417 pts
3 mentions across 1 sources
Reviewer notes
Auto-scanned review. These are observations, not a security certification.
Scored from trust signals (evidence-eval-v1): 42 GitHub stars; 1 contributors; last commit 98d ago; license MIT.
Things to check
- Scanned, not hands-on tested — this entry was auto-scanned from public metadata (GitHub metrics, license, security flags). No reviewer has run it, and no tool-specific limitations have been documented yet.
- Single maintainer. Consider the risk if this person stops maintaining the project.
How to evaluate tools before deploying →
Data shown here comes from public APIs and automated scanning. Reviewer notes reflect one person's experience. This is not a security certification or legal recommendation. Always evaluate tools according to your own organization's policies.
Evaluation
Scored from trust signals (evidence-eval-v1): 42 GitHub stars; 1 contributors; last commit 98d ago; license MIT.