[yaythomas]

Issue #, if available

None.

Description of changes

The AI review prompt reads CONTRIBUTING.md, but it focuses on code defects and requires a concrete code fix. Rule 9 instead requires reviewers to reject a broken approach without prescribing detailed code changes.

The prompt now checks rules 1 through 9, defines approach findings, reads required design information from the pull request description, and anchors every finding to an added line.

Related: JavaScript SDK PR #1104: have the AI review check the contributing rules.

Design

• Design note (rule 3): not needed. This changes no feature, option, or public API.
• Simpler option (rule 2): a focus bullet alone leaves the concrete-fix conflict unresolved.
• Root cause (rule 5): the prompt has no valid format for rejecting an approach.
• Net change: 0 source lines and 0 public API members. The prompt grows by 16 lines.

Assumptions

• Verified: both SDKs use the same shared workflow. It reviews the base checkout and requires added-line finding ranges.
• Unverified: how often Codex will report violations of judgment-based rules.

Decisions not covered by the issue or design note

• Approach findings rank below broken SDK invariants because invariant violations affect runtime behavior.
• The Python root AGENTS.md remains applicable because it contains contributor and replay requirements.

Checklist

• Every new branch has a test that fails without it (rule 4): no code branches were added.
• For a bug fix, the first commit adds a failing test (rule 5): this is not a code bug fix.
• Every edited test expectation is listed (rule 5): no test expectations changed.

Testing

• hatch run test:all .github/scripts/tests/test_ai_pr_review_workflow.py (2 passed)
• git diff --check
• hatch run python .github/scripts/lintcommit.py

The prompt takes effect after merge because reviews read the base revision.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[github-actions]

Codex AI review

No actionable findings. Static inspection only; the existing workflow tests do not validate how reviewers follow the added prompt instructions.

Reviewed commit df1c7f9eccb0c255a172c1fe09e70c3a0d6a97e9. Workflow run