Repolode
Submit
pskoett

pskoett/self-improving-agent

An OpenClaw skill that records learnings, errors, and feature requests so an agent can improve across sessions.

View on GitHubWebsite
Stars
776
Forks
116
Open issues
5
Last commit
4 hours ago

What is it?

self-improving-agent is a skill for OpenClaw, the personal AI agent framework. It gives the agent a structured way to capture learnings, errors, and feature requests during its work, so that lessons from one session can inform later sessions.

The skill lives in a dependency-free subfolder that ClawHub or a manual copy installs into an OpenClaw skills directory. An optional OpenClaw hook is included, and the project provides upgrade and uninstall guides. Captured entries are stored in a local .learnings/ directory, which the user owns and can review.

The repository also contains a Node.js development setup with tests, a TypeScript typecheck, and shell linting. The publishable skill itself does not require these tools. The project is OpenClaw-only; for other agents the author points to a separate multi-agent version.

Key features

  • 01Captures learnings, errors, and feature requests during agent sessions.
  • 02Keeps records in a local .learnings/ directory that the user can review.
  • 03Installs via the OpenClaw skill installer, ClawHub CLI, or a manual copy.
  • 04Includes an optional OpenClaw hook for automatic integration.
  • 05Ships with upgrade and uninstall guides.

Use cases

  • Helping an OpenClaw agent carry lessons from past errors into future sessions.
  • Tracking feature requests and recurring failures during agent-driven development.
  • Maintaining an auditable log of what an agent learned over time.

Quick start

  1. 1

    Install the skill into the active OpenClaw workspace with OpenClaw's built-in skill installer

    openclaw skills install @pskoett/self-improving-agent

Things to know

  • This version works only with OpenClaw; other agents such as Claude Code or Codex need the original multi-agent version.
  • The project has no license specified in its metadata, so reuse terms are unclear.

Topics

#agent-skills#openclaw#memory#knowledge-management#agent#personal-agent#openclaw-skill#openclaw-skills

FAQ

What is self-improving-agent?
An OpenClaw skill that records learnings, errors, and feature requests so an agent can improve across sessions. self-improving-agent is a skill for OpenClaw, the personal AI agent framework. It gives the agent a structured way to capture learnings, errors, and feature requests during its work, so that lessons from one session can inform later sessions.
Is self-improving-agent free to use? What license does it have?
self-improving-agent has no declared license. Its code is public, but without a license you have no explicit permission to reuse it; ask the author before relying on it.
Is self-improving-agent still maintained?
Actively maintained. The last commit to self-improving-agent was on Oct 22, 2026.
What can I use self-improving-agent for?
Helping an OpenClaw agent carry lessons from past errors into future sessions. Tracking feature requests and recurring failures during agent-driven development. Maintaining an auditable log of what an agent learned over time.
How do I install self-improving-agent?
Install the skill into the active OpenClaw workspace with OpenClaw's built-in skill installer: openclaw skills install @pskoett/self-improving-agent
Where can I get self-improving-agent?
The source code and installation instructions are on GitHub: https://github.com/pskoett/self-improving-agent

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects