cli-anythinglisted

## Full Role Description Transforms any GUI software into an agent-usable, stateful CLI harness using the software's native CLI interface as the backend. Agents get structured commands, JSON output, undo/redo, and a REPL — no GUI, no screenshots, no pixel-clicking. Trigger when: adding CLI capabilities to Blender, GIMP, LibreOffice, Inkscape, Shotcut, Audacity, or similar; automating a professional tool that has no API; building a CLI wrapper around a desktop application; controlling a software tool that only has a GUI; or generating a CLI harness from source code or documentation. Key capabilities: 7-phase pipeline (analysis → architecture → implementation → test planning → implementation → documentation → PyPI publish); direct invocation of the real software backend (ffmpeg, melt, LibreOffice, Blender); stateful session management with exclusive file locking and undo/redo; dual output modes (human-readable tables and machine-readable JSON via --json flag). Ideal for agents that need programmatic, reproducible control of creative and engineering tools. Also for: generating PyPI packages from CLI wrappers; building CLI tools for open-source software that only ships a GUI. # CLI-Anything Transform any open-source GUI application into a command-line tool an AI agent can operate — no display, no mouse. The agent uses Click subcommands to manipulate native project files (XCF, SVG, ODF, MLT XML, .blend) and calls the real software's CLI as a subprocess for rendering and export.