En

Agented

A command-line text editor designed for LLMs

开发流程 保存编码进度写业务代码切换开发任务优化Token消耗 通用 ★ 849 Updated 2026-08-02

Install & Use

Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:

Help me install this AI Skill: Agented.
It is used for: A command-line text editor designed for LLMs
Full Skill content: https://321skill.com/skills/agented-x-6/raw/index.md
Read that page and install it.

The prompt includes a link to the full Skill content. You can also view the full content.

This skill addresses the lack of history and concurrency control when AI coding agents edit files multiple times. In real-world development, AI agents need to modify multiple files in steps, but standard editors cannot persist edit states. Once context is lost, the process must start over from scratch. They also cannot detect if a file has been simultaneously modified by another agent, leading to overwrites or conflicts. Agented uses an SQLite workspace to store each edit version, building an undo tree and supporting transaction rollback, giving AI agents a complete edit history like humans have.

Usage is simple: just install the ae binary via brew, then run ae open <path> within a workspace to open a file. After that, you can use natural language instructions to perform operations like replace, insert, and delete. Each read operation returns a state_token. For write operations, pass the --expect parameter; if the file has been modified by another agent, the operation is rejected and new content is returned, ensuring no silent data loss. All edits are automatically versioned, allowing you to jump to any historical state at any time, or even branch into different edit paths.

It is ideal for AI coding agent scenarios requiring multi-step, multi-file editing. This is especially true for teams needing persistent edit history and support for concurrent collaboration—for example, multiple AI agents co-developing the same project, or a single agent completing complex refactoring over multiple sessions. It is also a perfect tool for writing business logic, switching development tasks, or saving coding progress, as it automatically manages state, freeing AI agents from manually tracking file change history.

It is recommended to use this tool uniformly when employing multiple AI coding agents or during long editing sessions. Note that it primarily targets plain text files and does not support binary files. Installation depends on Homebrew and SQLite, so ensure Homebrew is installed on your system. Short-form commands (e.g., ae s) are designed for AI agents to save token consumption; for human reading, the long forms are recommended.

Key Features

Unlike common editors like `cat`/`sed`, Agented supports state persistence, undo trees, and transaction rollback, with each write operation enforcing concurrency detection via `state_token` to prevent multi-agent edit conflicts. Unlike IDE-built-in editors, it is specifically designed for command-line LLMs, using short-form commands to save tokens and integrating into AI workflows without requiring a GUI.

Limitations

Supports only plain text files, not binary files; relies on SQLite for state storage, requiring Homebrew and a compatible macOS/Linux environment; edit history is stored locally in the `.agented/` directory, requiring manual synchronization for cross-machine migration.

FAQ

How do I install Agented?

Run `brew tap frane/tap && brew install agented` to install the `ae` binary command.

How do I undo an incorrect edit?

Use the `ae undo <file>` command. If ambiguity exists (multiple branches), branch information will be prompted. You can also specify a branch to revert with `ae undo <file> -b <branch-id>`.

Installation guide for AI assistants

If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:

Visit https://321skill.com/skills/agented-x-6/raw/index.md to read the original Skill definition (Markdown format) for Agented, and install it according to the instructions.