SkillClaw CLI

A research project that improves and deduplicates an agent's skill library from its sessions through a local proxy and an optional server

  • 2,691 ★ on the repository
  • Last commit
  • MIT
CLI

Broad access

Why this level

  • Proxies the agent's requests and rewrites its skills
  • With a shared server session data goes to shared storage
  • A research project with no promised results
All reasons and checks

amap-ml/skillclaw

Install

Manual install

skillclaw setup
skillclaw start --daemon

First install SkillClaw with the shell installer from the README (macOS, Linux) or by hand with Python (Windows).

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • Proxies the agent's requests and rewrites its skills
  • With a shared server session data goes to shared storage
  • A research project with no promised results

This is third-party code. Review the repository files before installing.

What it does

SkillClaw lets an agent's skills evolve from its own sessions. It has two parts. The client proxy listens on a local API (`/v1/chat/completions`, `/v1/messages`), intercepts the agent's requests, records session artifacts and manages the local skill library. The optional evolution server reads session data from shared storage, creates and improves skills and writes them back, and one server can serve several users and devices. Hermes and OpenClaw are supported, and since 20.04.2026 Codex and Claude Code with automatic proxy setup.

Who it is for. For researchers and agent enthusiasts who want to test the idea of automatically improving skills from real sessions.

Good fit when

  • You want to see how an agent's skills can be deduplicated and improved automatically
  • You need shared skill storage for several users and devices
  • You need an experiment with Hermes, OpenClaw, Codex or Claude Code on a local proxy

Not a fit when

  • You need a proven tool for a working project: SkillClaw is a research project
  • You cannot hand session data to shared storage

Example request

Run SkillClaw on a local proxy, work for a day as usual and show which skills it suggested merging

Limitations

This is a research project of the AMAP-ML organization, with no promised results, and the repository contains the paper skillclaw.pdf. The proxy proxies the agent's requests and rewrites its skills. If the shared evolution server is turned on, session data goes to shared storage. The shell installer targets macOS and Linux, and on Windows you set up Python by hand. After installation run `skillclaw setup` and `skillclaw start --daemon`. The repository was last updated on 17.08.2026.

How to disable. Stop the SkillClaw service and return the agent's proxy settings to the originals, then delete the installed files and the local skill library if you do not need it.

FAQ

What does SkillClaw do?

It improves and deduplicates an agent's skill library from its sessions.

What parts does it have?

A client proxy on a local API and an optional evolution server with shared storage.

Which agents are supported?

Hermes, OpenClaw, and since 20.04.2026 Codex and Claude Code with automatic proxy setup.

Where does session data go?

On the client proxy it stays local. If the shared evolution server is turned on, session data is read from shared storage.

Discussion

No questions or reviews yet.

CLI tools

A skill that distills a person's digital twin from chat logs and documents into a ready Agent Skill package

SkillBroad access1.1KRepository stars

A desktop app and CLI to install, sync and organize agent skills across Claude Code, Codex, Cursor and other tools

CLIExtended access6.3KRepository stars

A skill for creating skills that proves them by behavior: before-and-after runs, regression evals and ecosystem audit

SkillExtended access909Repository stars

A universal loader for Anthropic-format skills (SKILL.md) in any coding agent: install, sync and load on demand

CLIExtended access10.8KRepository stars
Editors’ pick

Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins

CLIBroad access391.9KRepository stars

A CLI and Python SDK to search, install, create and evaluate reusable agent skills from a shared library

CLIExtended access1.4KRepository stars

Comparison with similar tools

Entries that share the main search query, and how they differ.

EntryOfficialRepository starsMain difference
SkillClaw (this entry)no2,691
Agent Skill Managerno955a skill manager that scans before installing
upskillyes762a strong model builds a skill from traces and a weak one uses it
MCPMno1,012a package manager and registry for MCP servers

README in short

The README describes a project where skills evolve collectively: a client proxy, an optional evolution server, support for Hermes, OpenClaw, Codex and Claude Code. The repository has a paper. MIT license.

README badge

Report a problem

Foxx AISkillClaw
Foxx AI

I am Foxx AI and I have already vetted this tool. Ask about install, setup or anything else, and I will keep it simple.