odoo/documentation#21055
Created by Documentation, Felicia Kuan (feku)
Blocked
- Merge method
- Review
-
CI
- ci/documentation: (runtime 1467s)
- ci/codeowner_coverage: (runtime 0s)
- label
- odoo-dev:20.0-contributing-llms-style-rules-feku Runbot
- head
- 2d399a24fcdf9102d69906aa1c8d505c0abacc8f
| odoo/documentation | |
|---|---|
| 20.0 | #21055 ci/codeowner_coverage missing r+ |
| master |
[ADD] contributing: agent instructions and US style rules
Adds the files AI coding agents use when writing or reviewing docs
Context
Some R&D teams have started using AI tools to help draft documentation updates alongside their feature work. To support that, this PR adds a shared set of standards that AI tools can follow when writing or reviewing docs.
The goal is to make AI-assisted contributions consistent with our style guidelines from the start, so they need fewer revision cycles. The standards also ask AI tools to verify product claims against the author’s local Odoo instance, which makes contributions easier to verify. The technical writing team can then spend review time on clarity and usability rather than on fact-checking. This keeps quality consistent as contribution volume grows.
Summary of changes
AGENTS.md: repo instructions for agents. Tells agents which style rules apply to each file, which checks to run before handing the work back to the author, and how to scope PRs.CLAUDE.md: contains only@AGENTS.md. Claude Code loads CLAUDE.md automatically but not AGENTS.md, so this import gives Claude Code the same instructions without maintaining two copies.agents/style_rules.md: US team style rules, combining the US style guide with and the published guidelines by the BE team. US decisions take precedence where they conflict.agents/docs_ownership.md: which content paths are US scope, not US, or undecided.agents/pr_checklist.md: checklist the author copies into the PR description, including testing in a local Odoo instance on the matching version and adding screenshots.agents/pr_description.md: PR title and description rules, based on the team's existing PRs, with R&D task IDs and length caps added.