# sentient-agi/EvoSkill

EvoSkill — An open-source framework that automatically discovers and synthesizes reusable agent skills from failed trajectories to improve coding agent performance.

Repository: https://github.com/sentient-agi/EvoSkill
Canonical: https://ross.abutalabs.com/products/evoskill
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-24T23:18:46+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 89, longevity 13
- inputs: {"age_days": 182, "days_push": 9, "days_rel": 78, "gap_med": 9.5, "n_releases_24m": 5}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1142, forks 125 (observed 2026-08-28T04:03:44.547324+00:00)

## What it is
EvoSkill is an open-source Python framework that automatically discovers and synthesizes reusable skills from failed agent trajectories to improve coding agent performance. It extends GEPA-style feedback-driven optimization by jointly proposing skill and prompt mutations, evaluating variants on held-out data, and producing improved agent programs compatible with tools like Claude Code, Codex CLI, OpenCode, and OpenHands.

## Use cases
- automatically generate reusable skills for coding agents from failure data
- improve Claude Code or OpenCode performance on coding benchmarks
- evolve agent prompts and skills using any LLM provider
- run benchmark evaluations of agent variants on held-out tasks
- turn a general-purpose AI agent into a domain specialist

## When to choose
- you use a supported coding agent (Claude Code, OpenCode, OpenHands, Goose, etc.) and want automated skill/prompt improvement
- you have failed agent trajectories and want to mine them into reusable skills
- you want model-provider-agnostic agent optimization with benchmark-driven iteration

## When to avoid
- you need a general agent framework for building agents from scratch rather than optimizing existing ones
- your agent is not among the supported integrations
- you cannot run evaluation benchmarks or lack trajectory/failure data

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, llm-training, prompt-engineering, benchmarking, developer-tools
- domain: large-language-models, developer-tools
- platform: python, cli, cross-platform
- tags: skill-discovery, coding-agents, trajectory-mining, agent-evolution, claude-code, opencode, openhands, ai-agents, automation

## Member repositories
- sentient-agi/EvoSkill (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:44.547324+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:35:16.729692+00:00, confidence not recorded.
  - readme: https://github.com/sentient-agi/EvoSkill (fetched 2026-08-28T04:03:44.547324+00:00, sha be28eac2e563)
- Data as of 2026-08-30T08:39:29.467469+00:00.