strategic-revision

A Claude Code skill that turns peer review reports into structured revision roadmaps. Extracts every reviewer request as a separate task, maps dependencies as a directed acyclic graph, validates the graph computationally with NetworkX, and sequences work into execution blocks with co-author assignments and decision points.

Trigger

Invoke with /strategic-revision or any request to create a revision plan with DAG validation from peer review reports.

Installation

1. Download the strategic-revision.skill file from this repository
2. In Claude Code, go to Settings → Capabilities → Skills → Add and upload the .skill file

Input Requirements

The skill requires two inputs, placed in the working directory or provided when prompted:

1. Peer review reports — verbatim reports from each reviewer and the editor. Any format (PDF, Word, plain text). Provide the full unedited text; the skill traces every task back to a specific reviewer quote, so summaries are insufficient.
2. Original manuscript — the submitted version under review. The full manuscript is preferred; at minimum, the abstract and section structure are needed for dependency mapping.

What It Does

Analyzes reviewer comments and the manuscript to produce a structured revision roadmap with:

• Atomic task extraction — parses each reviewer comment into separate actionable tasks with verbatim quotes, preventing details from being overlooked
• Classification by type (empirical, argumentative, structural, clarification, editorial)
• Dependency mapping as a directed acyclic graph, catching sequencing errors before work begins
• Computational validation — acyclicity checks, parallel batches, critical path analysis, and bottleneck identification (surfaces tasks that block the most downstream work)
• Execution blocks with GO/NO-GO decision points and explicit co-author sync points

Workflow

Files

strategic-revision/
├── SKILL.md                              # Skill definition and workflow
├── scripts/
│   └── dag_validator.py                  # NetworkX validation script
└── references/
    ├── phases.md                         # Phases 1-5 (including 3b) instructions
    ├── phase6-dag-validation.md          # Phase 6 optimization instructions
    └── task-schema.md                    # JSON schema for revision_tasks.json

👉 Claude Code users! Requires Python 3 and networkx (pip install networkx).

Limitations

• No effort estimation. Tasks are sequenced by dependency, not by expected effort. A task requiring two hours and one requiring two weeks are treated equally.
• No strategic pushback guidance. The skill extracts and organizes all reviewer requests but does not flag candidates where “we respectfully disagree” may be the appropriate response.
• Dependency mapping requires user judgment. The tool validates DAG structure but does not generate dependencies automatically. Different users may construct different graphs from the same reviews.
• No response letter integration. The plan organizes tasks but does not generate response letter text or track task-to-reviewer mappings for the letter.
• Static plan. If early tasks change results substantially, the plan does not update automatically. Replanning is manual.
• Empirical paper template. The A-to-E execution block structure fits empirical accounting and finance papers. It may not suit qualitative, theoretical, or interdisciplinary work without adaptation.

Intended Use

Designed for initial planning and co-author coordination at the start of a revise-and-resubmit. Not a substitute for execution tracking or adaptive replanning. A practical workflow: use this skill to produce the revision roadmap, then migrate to a simpler checklist for daily work.