Agented

This Skill addresses the lack of state persistence, version history, and concurrency control when AI agents edit code or text. In real-world development, AI agents need to edit multiple files across multiple invocations, but standard tools like cat/grep/sed cannot retain context, leading to redundant work and risk of conflicts. Agented stores state in an SQLite workspace, with every edit recorded in an undo tree that supports branch navigation, ensuring no modification is lost.

Usage is simple: just open a file via ae open <path>. It returns a state_token, which must be included in all subsequent write operations (e.g., ae s for replace, ae i for insert). If the file is modified by another agent, the write operation is rejected (exit code 3) and returns the new content and token—simply retry. Supports multi-line content via pipe input, e.g., cat patch.txt | ae s foo.go -r 12:14 -i. Commands use short forms by default to save tokens, such as ae s foo.go -r 12:14 -w "..." -x ab12cd34.

It is ideal for AI agents and agent development teams that require safe, multi-step file editing, especially in scenarios involving frequent refactoring of legacy code, root cause debugging, or synchronizing project context. If you use tools like Claude or Cursor for complex coding tasks and want to preserve a complete edit history, Agented significantly reduces the manual overhead of version management.

We recommend using this tool consistently when you need edit history to persist across sessions. Note: it primarily targets text files and does not support binary files; it depends on the ae binary, installable via Homebrew (brew tap frane/tap && brew install agented); for simple one-off edits, using the platform's built-in editor is more efficient.