ArchestackBest of Open-Source AI: List on GitHub ↗ (opens in a new tab)

#2727th of 28 in Memory

skillclaw

Local proxy that evolves reusable agent skills from recorded sessions

Repository on GitHub ↗ (opens in a new tab)
Stars
2.7k
License
MIT
Last commit
Oct 2026
Language
Python

Missed the latest GitHub check; these facts may be older.

Overview

SkillClaw runs a local proxy exposing /v1/chat/completions and /v1/messages between an agent and its LLM. It records sessions and manages a local SKILL.md library. An optional evolve server reads sessions from shared storage (local, Alibaba OSS or S3), then deduplicates and rewrites skills with a workflow or OpenClaw-driven engine. It integrates with Hermes, Codex, Claude Code and OpenClaw.

Who it is for: Users of Hermes, Codex or Claude Code who want their skill libraries curated

Strengths

  • Works with Hermes, Codex, Claude Code, OpenClaw, QwenPaw and any OpenAI-compatible API
  • Client-only install works alone; the evolve server is optional and can be added later
  • Shared storage backends: local filesystem, Alibaba OSS, S3, with optional Nacos for skills
  • Ships doctor/restore commands and a local dashboard for skills, sessions and version history

Weaknesses

  • Automatic evolution needs a separate evolve server process plus an upstream LLM endpoint
  • No Docker or compose files; install is from a git clone with a shell script
  • Windows install is manual; no native PowerShell installer
  • No tagged releases; Hermes session boundaries fall back to proxy heuristics without session headers

What it needs

  • no GPU
  • Needs Python 3.10+, OpenAI-compatible API or AWS Bedrock, OpenClaw (only for agent engine), Alibaba OSS or S3 (optional, for sharing)
  • Models: OpenAI-compatible APIs, AWS Bedrock
  • port 30000

Also in Memory

See all 28
Also in Memory
RankProjectScore
1hindsightAgent memory server with retain, recall and reflect operations60.6k stars+12,039 (+25%) in 30 daysMIT95 out of 100
2gbrainMemory store with sources and a knowledge graph for coding agents32.7k stars+2,034 (+6.6%) in 30 daysMIT91 out of 100
3MemPalaceLocal verbatim memory store for AI agents with semantic search and MCP60.5k starsMIT90 out of 100
#4SupermemoryMemory and context engine for AI agents, with user profiles and hybrid search31.6k starsMIT89 out of 100
#5Mem0Long-term memory layer that stores and retrieves user facts for LLM agents Live demo ↗ (opens in a new tab)68.9k starsApache-2.088 out of 100