We have a lot of rules for writing docs. Agents don't follow those. So I created a skill:
2 Likes
I'd be interested to read feedback of its responsible usage, where I define "responsible" as, but not limited to, the following points.
- Disclose the usage of AI.
- Take responsibility for the output, including reviewing and validating the output for accuracy and ensuring it resolves an issue.
- You must check the AI's terms of use, and ensure that outputs are not reconstructed from copyrighted sources.
- The output must be reproducible. That means Include the name and version of the tools, and the inputs provided to generate the output. An AI skill such as this one does exactly that. Publishing this skill on GitHub also allows other people to contribute to improve and refine it after a mere mortal performs Item 2.
The greatest problem in Plone documentation remains Item 2, with or without AI. I've seen approvals of Plone documentation pull requests where trusted knowledge experts approved things that are technically incorrect or obsolete. Sure, mistakes and omissions happen, especially when reviewing a PR of over 1000 lines or dozens of files. But when it's obviously inaccurate to me, a relative newcomer to Plone with almost no practical experience developing it, I wonder what their review process was that allowed the errors to pass through.