Workspace · Revision toolkit
996102e0
Peer-review revision toolkit
Use this workspace when a manuscript has entered peer review and the revision now mixes prose changes, new analyses, changed endpoints, corrections, and final archive work. It routes each task to the smallest relevant guide or worked example while keeping the submitted record recoverable.
8 / 8 audit followed by a reviewer-requested 4 / 4 sensitivity amendment remains 8 / 8 submitted + 4 / 4 post-review, not “12 pre-specified analyses.”
Choose what you need to do now
Classify the reviewer request
Decide whether the item is a clarification, correction, sensitivity amendment, analytical amendment, endpoint amendment, or measurement amendment before changing the record.
Run and preserve new analysis
Declare the post-review specification space, record whether outcomes were already seen, keep valid failures in the denominator, and save the amendment separately.
Write the response letter
Bind each reviewer item to its category, timing, action, evidence, manuscript location, archive location, and execution denominator.
Record what changed
Give every material v1 → v2 change a stable ID and distinguish editorial edits from corrections and analytical amendments.
Build the revision package
Use the real CLI to scaffold response, change, amendment, software-identity, and editor-facing evidence-map records, then validate the structural manifest.
Run the final handoff check
Verify reviewer-item closure, temporal denominators, response/manuscript agreement, software identity, archive evidence, and unresolved limitations before resubmission.
Request → record → evidence map
| Reviewer situation | Preserve | Do not do | Continue with |
|---|---|---|---|
| Wording is unclear; no analysis changes | response item + manuscript location | invent an analytical amendment | Response-letter guide |
| Reviewer asks for another threshold or exclusion | dated amendment + separate denominator | merge it into the submitted denominator | Reanalysis example |
| Reviewer asks for a different endpoint | endpoint-specific amendment record | mix endpoints inside one robustness denominator | Amendment guide |
| One valid post-review branch fails technically | valid denominator + failure reason + incomplete status | relabel 3 / 4 as 3 / 3 |
Revision-package example |
| Review exposes a coding/import/analysis defect | superseded evidence + defect record + regenerated outputs | delete the superseded record | Version-change manifest |
| All reviewer work is complete | response matrix + change manifest + evidence map | rely on the response letter alone | Resubmission readiness |
The minimum revision record
A reconstructable review round should answer seven questions without oral explanation:
- What existed at submission?
- Which reviewer item introduced this change?
- Had the relevant outcomes already been inspected?
- What analysis or documentation action was taken?
- What is the submitted denominator and what is the post-review denominator?
- Where did the manuscript change?
- Which archived file supports the response?
If one of those answers is hidden in email, lab memory, or an untracked local folder, the revision record is not yet independently reconstructable.
Fast executable path
Create the governed scaffold:
gazeaudit-revision-package init \
--output-dir revision-package \
--project-slug my-study \
--review-round 1
Populate the response matrix, version-change manifest, amendment folders, and editor-facing evidence map, then run:
gazeaudit-revision-package validate --root revision-package
A passing validation establishes structural provenance only. It checks the expected package files and deterministic manifest integrity; it does not certify the scientific validity of the analysis or the editorial quality of the manuscript.
Use the revision-package quickstart for the shortest end-to-end exercise. For the full break-and-repair exercise, use the revision reproducibility package worked example.
Three records that should not be collapsed
The endpoint, decisions, valid denominator, successes, failures, software identity, and claims that existed before peer-review amendments.
Reviewer-triggered clarifications, sensitivity analyses, analytical changes, endpoint additions, measurement changes, and corrections, each with timing.
The revised manuscript, response matrix, version-change manifest, evidence map, software record, limitations, and publication-associated archive.
The final archive can contain all three layers, but they should remain identifiable as different temporal records.
Worked routes
One sensitivity amendment
Use the reviewer-requested reanalysis example when the main challenge is adding one post-review analysis without rewriting the submitted declaration.
Multiple reviewer items
Use the revision response package example when clarification, new analysis, and endpoint changes must all agree across the response letter, manuscript, and archive.
Fast package creation
Use the revision-package quickstart when you want the shortest real-CLI path from a synthetic reviewer request set to a structurally validated evidence package.
Executable package validation
Use the reproducibility-package CLI example when you want to create the actual scaffold, deliberately break a governed file, see the validator fail, and repair it.
Complete submission → final record
Use submission-to-accepted-record to trace the synthetic submitted 8 / 8 record, reviewer-requested 4 / 4 amendment, separate endpoint amendment, response matrix, version manifest, final evidence map, and publication-associated archive.
Five-minute pre-resubmission check
Before the revised manuscript leaves the team, verify that:
- every reviewer item has a stable disposition;
- every new analysis is labelled with its post-review timing;
- submitted and post-review denominators are not collapsed;
- failed valid branches remain visible;
- endpoint amendments remain endpoint-specific;
- corrections retain superseded evidence and affected-claim provenance;
- response wording matches manuscript wording;
- every material response points to real archived evidence;
- the final software identity is recorded;
- the evidence map points from each final claim component back to its temporal evidence layer.
Then run the fuller resubmission readiness and editor handoff gate.
Evidence boundaries
GazeAudit’s frozen empirical case-study outcomes are unchanged and remain authoritative only within their own protocols: GazeBase incomplete, Korthals robust_negative, and Pedrotti/de Chambrier materially_fragile. The reviewer examples linked from this page are synthetic teaching material.
Full review route
For the complete interpret → draft → reconstruct → amend → respond → archive sequence, continue to the manuscript review path. For durable publication provenance after the revision is finalized, continue to the reproducible publication workflow.