How to Contribute to AI4L

First off, thanks for taking the time to contribute!

Your feedback & discussions are always welcome!

Just reach out to us via GitHub discussions, and we will be happy to chat about your ideas, suggestions, or questions.

Below, you will find a brief guide to evolving the AI4L prompt and customizing it for your needs.

Customizing the Expert List

The list of experts prioritized for the "Recommended Reading" section can be modified by changing AI4L.md (9.5)

Use a CLI Environment
Agents & skills are located in the .claude / .opencode folder
After modifying files, use "/check" to check all AI4L files for inconsistencies or errors. The skill should be auto-loaded by the CLI.
The CLI might need a /clear or even an exit from the CLI to reload the changed agent or skill code

Use /er create <topic> to create an evidence review
Use /er audit to audit the review
Inspect the resulting QA.md for any errors, inconsistencies, or misbehavior of the auditing AI.
Fix issues by improving the audit instruction at the beginning of "AI4L.md" and throughout the document as inline comments for the auditor.
This can be done interactively, in dialogue with the AI itself, by asking it how to improve the prompt to avoid a particular mistake or misunderstanding in the future.
Repeat until satisfied
Proceed to step 2 ONLY WHEN repeatedly happy with the behavior of the QA process itself.

When unhappy with the actual content of the ER, interactively work with the AI on how to improve / change / add checklist items to steer the generating AI in the desired direction.
Remember the reverse, QA based approach as described in README.md
QA items work best when they are not only negative (DOES NOT ...) but also clearly express what is expected and give a precise example.
Repeat the cycle " Create > Audit > Improve audit checklist items > ... " until satisfied

Check out our Lessons Learned document for insights and best practices.