Summary
As a reviewer, I am finding it increasingly obvious that the clankers aren't behaving themselves :)
I'd like to ensure that if this project is using AI, that all LLMs contributions are using are going to align on common ground. I'm advocating for the creation of an AGENTS.md file describing the project, its structure, and its first instruction. That being, to the tune of:
Before working on this project, study .agents/feedback/*.md files which provide ground rules for contributing.
Ground rules could be:
- Code formatting rules, DRY, good naming conventions etc
- Licence and code attribution
- Update documentation when new features are added
- Never present assumptions as facts; always verify from source before making claims
Formation of a CONTRIBUTING.md with the following to be included in PR descriptions:
- Whether an LLM was used
- If so, state it. State which one, and which model
- What was the LLM used for
Why
LLMs are like messy junior developers at the best of times, and at the worst of times, totally unfit for purpose. At least provide them with the contextual knowledge to be somewhat helpful. Also, provide transparency for peer review to happen accordingly.
Summary
As a reviewer, I am finding it increasingly obvious that the clankers aren't behaving themselves :)
I'd like to ensure that if this project is using AI, that all LLMs contributions are using are going to align on common ground. I'm advocating for the creation of an
AGENTS.mdfile describing the project, its structure, and its first instruction. That being, to the tune of:Ground rules could be:
Formation of a
CONTRIBUTING.mdwith the following to be included in PR descriptions:Why
LLMs are like messy junior developers at the best of times, and at the worst of times, totally unfit for purpose. At least provide them with the contextual knowledge to be somewhat helpful. Also, provide transparency for peer review to happen accordingly.