Proposal: Add `pyproject.toml` to the repository, i.e., an env

[VascoSch92]

The repo has 13+ Python files across skills/ and tests/, uses third-party packages (requests, pyyaml, pytest), and has a .gitignore that already accounts for Python artifacts. However, dependencies are completely implicit.

Since our goal is to increase plugins and skills adoption, I propose adding a pyproject.toml file to standardize the project configuration.

Pros:

• Reproducible environments: Right now, contributors have to read through the source to figure out what to pip install.
• Tests: we could have more easy-to-run and better tests
• Linting, formatting and static type check (they prevents bugs and errors)
• Python version requirement: Some scripts use from future import annotations and dataclasses, signaling a minimum Python version.

If you think it would be useful, just say the word and I'll do it! ;-)

[enyst]

That’s a good question! I wonder what other skills repos out there (e.g. anthropic’s but maybe not only) are doing on this.

We could look at it alternatively like:

• IMHO this is not for contributors, it’s for agents;
• it’s not python, it’s any language the skill happened to have (we just merged a skill with npm script) 🤔

I guess in a way I still look at these scripts as examples (“the agent will figure out details”), but maybe I’m wrong.

Actually they are run inside the agent-box, inside the agent’s workspace at runtime, so they should have the same requirements if any, that the SDK has. I idly wonder if there’s a better way than to duplicate those here… WDYT?

[VascoSch92]

I checked and antrophic skills package doesn't have a pyproject.toml.

My point was that as we add more scripts and people begin using these skills, maintaining quality and support could become difficult.

However, it might be a little too early to worry about that just yet.