Skill Detail

Evolve reusable coding-agent skills from failed trajectories with EvoSkill

Mine failed agent runs for reusable skills, benchmark the candidates, and keep only the variants that improve a supported coding agent over your baseline.

Templates & WorkflowsMulti-Framework
Templates & Workflows Multi-Framework Security Reviewed
⭐ 489 GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill evolve-reusable-coding-agent-skills-from-failed-trajectories-with-evoskill Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Python environment, a supported coding agent such as Claude Code, OpenCode, OpenHands, Goose, or Codex CLI, and a benchmark or task set to optimize against
Install & setup
Follow the repository Quickstart to install EvoSkill, connect a supported coding agent, and run the evaluation loop against your chosen benchmark or task set.
Author
Sentient
Publisher
Organization
Last updated
Apr 16, 2026
Quick brief

Use EvoSkill when the job is to turn repeated coding-agent failures into reusable skills instead of fixing prompts by hand one task at a time. The operator workflow is narrow and publishable: point EvoSkill at a supported coding agent and task set, induce candidate skills from failures, evaluate them, and keep the versions that measurably outperform the baseline.

How it works

What this skill actually does

The scope boundary is what saves this from being a generic product card. This is not a broad agent framework listing or model marketplace entry. It is a concrete failed-trajectory-to-skill-improvement loop for coding agents, with benchmarking and selection built into the workflow.