Review a BRD paragraph by paragraph, capture clarification questions for ambiguous or incomplete requirements, and generate a single .docx with Word comments plus tracked revisions.
Prefer the bundled pipeline so the final deliverable is a Word-native review artifact instead of chat-only notes.
.docx path.```bash
python scripts/brd_review_pipeline.py init-review \
--input
--output
```
paragraphs[] item, keep paragraph_index, style_id, heading_path, and source_text unchanged.needs_comment to true when the paragraph is unclear, incomplete, internally inconsistent, or missing acceptance criteria, data definitions, ownership, dependencies, assumptions, or edge cases.comment_question as a concise reviewer question suitable for a Word comment.needs_revision to true when the paragraph should be rewritten for precision, completeness, grammar, or testability.proposed_replacement as full replacement language, not fragments.issue_tags to make the reason explicit. Prefer tags such as ambiguity, scope, actor, data, workflow, exception, dependency, acceptance-criteria, nonfunctional, term-definition, or conflict.```bash
python scripts/brd_review_pipeline.py materialize \
--input
--review-json
--output
--author "Codex BRD Reviewer"
```
needs_comment=true.needs_revision=true.$doc skill workflow to render and visually inspect the result before delivery..review.json .reviewed.docx .docx is the primary deliverable.references/review-json-schema.mdscripts/brd_review_pipeline.pyInstall once if missing:
python -m pip install python-docx lxml
共 1 个版本