review-processing

name: review-processing description: Process raw review comments into a drafting-ready revision package

Convert raw reviewer comments into a structured revision package that separates manuscript changes from rebuttal-only, deferred, or declined items.

NEVER assume every reviewer comment must be incorporated into manuscript text
Always preserve stable issue IDs across all generated artifacts
Explicitly separate manuscript edits from rebuttal-only, deferred, and declined items
Flag requests requiring new analysis or new data instead of disguising them as straightforward writing changes

Parse editor letter and reviewer blocks into atomic issues
Assign stable issue IDs by source and comment order (for example R1-C3, AE-C1)
Normalize each issue into a concise actionable summary
Classify each issue by content area, severity, and evidence requirement
Assign a disposition for each issue: revise_manuscript, rebuttal_only, defer, or decline
If existing_manuscript_file is available, map manuscript-relevant issues to target sections and likely edit scope
Build review_decision_log.md as the authoritative triage ledger
Build review_comment_map.md containing only issues marked revise_manuscript
Build revision_brief.md summarizing revision priorities, cross-cutting changes, and section-level revision agenda
Build revision_open_questions.md for high-risk declines, deferrals, and unresolved strategic choices requiring user approval

If review_file is missing or unreadable, request a valid path or pasted comments and pause
If parsing reviewer blocks fails, switch to manual segmentation and keep stable issue IDs
If existing_manuscript_file is unavailable, produce revision artifacts without section routing and flag mapping as incomplete
If multiple high-severity issues imply conflicting revision strategies, surface them in revision_open_questions.md and stop for user confirmation