chekos
Providing a prompt file for using coding automation tools
First off, thank you all for the fantastic work on v2 of this library!
I'm wondering if, going forward, people will use not only MCP for creating and automating Datawrapper charts, but also coding automation tools like OpenAI Code, Anthropic's Claude Code, or Cursor.
To make this work, a prompt file could help teach these tools how to write and execute the wrapper code properly. The challenge is that whenever a new feature gets added or something gets improved, the prompt file would need updating alongside the documentation. So it becomes a maintenance question as well.
Is there a chance to provide such a prompt file? Recent examples can be found here.
https://github.com/dontriskit/awesome-ai-system-prompts
Thanks for sharing this! This looks interesting. Have you taken a look at @palewire's datawrapper-mcp? https://github.com/palewire/datawrapper-mcp I think that solves what you're talking about.
Yes, I found out about it today - thanks for pointing it out! The MCP solution is excellent for what it does. However, there's an interesting distinction here: MCP operates through a server-based protocol, while coding automation tools like Claude Code or Cursor work directly with the user's local filesystem and Python interpreter - they also need different prompting.
For example, when using Claude Code or Cursor, you're often working on Python scripts that will be part of larger automation pipelines or scheduled jobs. A prompt file would help these tools understand the library's patterns and best practices.
Also it's kind of a workflow issue, the prompt file needs to be closely tied into the main repo and needs and update when major code updates are merged (just like the readme/docs do). But I totally understand if this is more of a side quest regarding the main library.
I'd be happy to contribute an initial version, but it would need to become part of the library's maintenance cycle.
This is interesting, I understand what you're saying.
I think it'd be a great addition. You could send a PR and we'll take a look! Thanks in advance!
I've used context7 for this - it'll give me up to date documentation on libraries. I assumed it depended on the quality of the docs and codebase documentation but this makes more sense
Added an AGENTS.md as a playbook. It includeds golden path for common tasks, do/don’t rules, data-shape expectations, and an exact output format. I hpe it makes agent-based usage predictable and reproducible. I would expect there are still lots of aspects to be improved but it's a first draft that hopefully makes clearer why this would be useful in the future.