Skip to content
Get packs

A-Evolve Agent Improvement

Diagnose agent failures from run logs and turn recurring errors into targeted skills, prompt patches, and knowledge entries.

Source Download skill
Repository
aiming-lab/AutoResearchClaw
Stars
14,802
Licence
MIT
Files
1 file

What it does

Once installed, the agent follows a five-step loop (Solve, Observe, Evolve, Gate, Reload) to review an AI agent's past runs, find recurring failure patterns, and propose fixes. The fixes are saved as new skill files, short system prompt additions, or JSON knowledge entries that future runs can load. It is written as instructions only: the agent reads your logs and writes proposed artifacts, and it does not change tool code unless you explicitly ask.

Great for

  • Developers running AI agent experiments who want a structured way to learn from repeated failures
  • Researchers using the AutoResearchClaw pipeline who want to improve weak stages
  • Teams keeping a running log of what changed in an agent's prompts and skills over time

Skip it if

  • One-off errors or cosmetic issues, which the skill itself says should be ignored, since generating fixes for them adds noise
  • Production systems where prompt or skill changes are applied without human review, because the skill's gate step is only a judgment check and mistaken edits could cause regressions

Try asking

  • “Diagnose why my agent failed 12 of 50 tasks in last night's run and group the errors by cause”
  • “Generate a targeted skill file for the API pagination errors that keep recurring in my logs”
  • “Evolve my AutoResearchClaw research pipeline using the latest artifacts/rc-* outputs and reviews.md”
  • “Propose a minimal system prompt patch for the citation format mistakes in my batch 2 results”
Step-by-step guide: how to use A-Evolve Agent ImprovementInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    a-evolve.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip a-evolve.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip a-evolve.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give it real run artifacts and pass/fail counts rather than asking for general advice, since the method depends on observed evidence
  • Review every proposed skill, prompt patch, or knowledge entry before placing it in your project, because the agent generates these and you are responsible for accepting them
  • Keep the three-skills-per-cycle limit and small mutations the skill recommends to avoid broad, risky rewrites

What's inside

  • SKILL.md

Redistributed unmodified from aiming-lab/AutoResearchClaw under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.