Self-Improvement Evaluation

SkillDev tools

Independent evaluation protocol for bounded self-improvement candidate proposals.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Self-Improvement Evaluation skill

What this skill tells your AI

The instructions your AI receives, as published by abivan-tech/opencode-agentic-workflows in .agents/skills/self-improvement-evaluation/SKILL.md and read by ahel’s review.

Use this skill when evaluating a bounded self-improvement candidate.

Evaluation Inputs

  • candidate manifest
  • baseline manifest
  • current promotion policy
  • synthetic scenario manifest
  • optional held-out private scenarios

Required Output

Return one of:

  • PASS_AUTO
  • NEEDS_HUMAN
  • REJECT

Evaluation Dimensions

  • reliability/regression
  • safety/governance
  • cost/latency
  • reuse
  • routing quality

Rules

  • Keep the evaluator independent from the author.
  • Use frozen baseline and candidate versions.
  • Keep metrics normalized and minimal.
  • Never serialize full transcripts or secrets.
  • Fail closed on policy violation, budget breach, or insufficient evidence.

Signals

GitHub stars
28
Forks
3
Last commit
Jul 2026
Advanced
Catalog kind
skill
Gateway key
self-improvement-evaluation
Source
github.com/abivan-tech/opencode-agentic-workflows