Classic confinement request for quillwright

  • name: quillwright

  • description: Quillwright is a desktop writing app for novels and other long documents (MIT license, written in Rust). A project is a plain folder of Markdown files. It is designed to run next to AI command-line agents (Claude Code, Codex, Ollama, and similar) that edit the same folder: it watches the folder, reloads external changes safely, and has an assistant panel that sends the text to the AI tool the user already has, then shows the answer as a diff for the author to accept or reject.

  • snapcraft: quillwright/snap/snapcraft.yaml at main · fabiocmazzo/quillwright · GitHub

  • upstream: GitHub - fabiocmazzo/quillwright: A distraction-free writing environment for novels and long-form fiction, built on plain Markdown and designed to work alongside AI agents · GitHub

  • upstream-relation: I’m the upstream author and maintainer (GitHub: fabiocmazzo), and I publish the snap myself. It is built from the repository by GitHub Actions.

  • supported-category: AI agent/assistant that runs arbitrary user-directed code and needs to reach files and programs that aren’t known at build time.

  • reasoning:

    The assistant panel runs a command that the user chooses in the settings: their AI command-line tool (claude, codex, ollama run <model>, or any other command). Quillwright sends the prompt on stdin, reads stdout, and shows the answer as a diff. The lint and format tools work the same way (a user-configured command, rumdl by default).

    These programs are not known at build time. They are third-party tools that the user installs and updates on their own, in places such as ~/.local/bin, npm or nvm prefixes, or /usr/local/bin, and they need their own configuration and sign-in. They can’t be shipped in the snap, and as far as I know no interface lets a strictly confined snap execute arbitrary host binaries.

    Projects are ordinary folders that the user and their agents edit at the same time. The app watches them recursively and saves atomically (a temporary file plus a rename in the same folder). Quillwright itself does not read dot files or anything outside the projects; on its own, project access could be covered by home and removable-media. Running the user’s programs is what requires classic.

    How the snap behaves under classic:

    • base: core24, with GTK 3 and Mesa bundled. The launcher starts the binary through core24’s dynamic loader with --library-path pointing at the snap’s and core24’s libraries, so it never loads host libraries.
    • Before starting a host program, the app removes the environment variables the launcher set (library, GTK, and Mesa paths) and restores the host’s XDG_DATA_DIRS, so those programs run in their normal environment.
    • The app itself never connects to the network and has no telemetry.
    • CI builds the snap, installs it with --classic, and starts it under Xvfb on every change to the packaging.

    A revision (0.4.0) is uploaded and waiting for this review; 0.5.0 is released upstream and will follow.

I understand that strict confinement is generally preferred over classic.

I’ve tried the existing interfaces to make the snap to work under strict confinement.

I went through the interface list: home and removable-media would cover the project folders, but none of them allows running the user’s own programs from the host, which is the app’s main feature. I’m happy to test any interface you suggest.

This request has been added to the queue for review by the @reviewers team.

Please see the template on this page:

Thanks! I’ve updated the first post to follow the template.

1 Like

Hello @fabiocmazzo :slight_smile:

It is plausible that there are technical reasons why quillwright needs classic to work properly in all scenarios, but please note that classic confinement is a sensitive matter and it is reserved for mature, well-known applications published by mature, well-known entities. As of today, I believe that quillwright doesn’t meet these criteria because of the following reasons:

  • The project seems to be very fresh, according to the upstream repository.
  • The projects seems to have little/none community around according to upstream repository (contributors, issues, PRs, etc.).
  • I could not find evidences that the project has a strong enough user base at the moment.

Thus, considering these factors, I think quillwright should not get classic confinement as of now.