sandroandric

AgentHandover

Name: AgentHandover
Author: sandroandric

Development community intermediate

SOURCE https://github.com/sandroandric/AgentHandover

Description

What if OpenClaw, Claude Code, Codex etc. knew how to do your work without you saying it? AgentHandover observes you, learns and teaches your agents with self-improving skills.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

AgentHandover

Work once. Hand over forever.

License

Demo · Quickstart · Use Cases · What a Skill Looks Like · How It Works · Connect Your Agent · Install · Privacy

The Struggle - The Discovery - The Transformation

Demo

New demo · Original demo

Show it once. Hand it off forever.

AgentHandover watches how you work on your Mac, turns your workflows into reusable **Skills**, and lets agents like **Claude Code**, **OpenClaw**, **Hermes**, **Codex**, or any MCP-compatible tool execute them the way you do it.

Each Skill captures the *what*, the *why*, and the *how* — steps, strategy, decision logic, guardrails, and your writing voice. And they're **self-improving**: agents report back after every execution, successes boost confidence, deviations become new decision branches, failures trigger corrections.

You already know how to do your work. Now your agents can too.

Why this matters

Getting an AI agent to do real work today means writing prompts or hand-crafting skills — brittle, time-consuming, and stale the moment your process changes. Worse, those skills capture *what* you do, not *how* you decide.

AgentHandover flips it. Instead of telling the agent how you work, you just work. The system watches, infers the strategy behind your clicks, and produces Skills with the *why* built in — selection criteria, guardrails, decision branches, your voice. Because agents report back after every execution, Skills get better the more they're use