Per-file review against CONTRIBUTING.md and DESIGN.md

Runs workflow.py (next to this file) with run_workflow: one read-only lite child session per changed text file, then one session that removes duplicates, drops findings that are clearly wrong, and sorts by severity (critical → info). The final list is in pt-BR, printed in the run output between FINAL_MARKDOWN_BEGIN and FINAL_MARKDOWN_END.

Choose what to review

Reviewers check out HEAD from origin, so the commits must be pushed.

Constants at the top of workflow.py:

• HEAD_REF (default "HEAD"): the tip to review. Check out the branch first, or set a ref/SHA.
• BASE_REF (default None): None uses the merge-base with origin/develop. If the checkout is develop itself, it uses HEAD^1 instead, which is the last merge.
• REPO_DIR (default /home/ubuntu/repos/boto): the local clone used to list the files.

Edit the constants in a copy of the file, not in the repo (for example cp workflow.py ~/workflows/review.py). Then pass that path as script_path.

Deleted files and binary files (images, fonts) are skipped. Each changed text file gets one reviewer. Tell the user how many files that is before starting a wide range.

Run

1. git -C <REPO_DIR> fetch origin, then check out the target.
2. run_workflow(workflow_name="review-contributing-design", script_path=<path to workflow.py>).
3. When it finishes, save the text between the FINAL_MARKDOWN_* markers to a .md file and attach it for the user. Spot-check one or two findings before reporting.