paper-audit

Deep-review-first academic paper audit for LaTeX, Typst, and PDF documents. It is a reviewer and gate-checking workflow, not a source editor.

Use It For

• Quick submission-readiness screening.
• Final-week mechanical checks for em dashes, AI-tone vocabulary, abstract result gaps, citation/label/equation hygiene, and paragraph-shape weak signals.
• Reviewer-style deep critique with major/moderate/minor findings.
• PASS/FAIL gate decisions calibrated for submission blockers.
• Re-audits that compare a revision against a previous audit.
• Review workspaces with traceable artifacts, claim maps, quote checks, and revision trajectories.

Do Not Use It For

• Direct .tex or .typ source editing as the first step.
• Compilation repair as the main task.
• Paragraph-level polishing without an audit goal.
• Free-form literature survey writing.
• Cover-letter generation or claim alignment; use cover-letter.

Mode Router

Mode	Use when	Primary command
quick-audit	You want a fast script-backed readiness screen	uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode quick-audit
deep-review	You need reviewer-style findings, workspace artifacts, and a revision roadmap	uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode deep-review --focus full
gate	You only care about hard submission blockers	uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode gate --venue ieee
polish	You want precheck-only handoff before style editing	uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode polish
re-audit	You have a previous report and need regression comparison	uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode re-audit --previous-report report_v1.md

Compatibility aliases: self-check -> quick-audit; review -> deep-review.

Deep-review dispatches five committee agents and six or more lane agents, then uses synthesis_agent.md. Mode-specific agents cover gate, re-audit, and post-consolidation revision suggestions. Specialized reviewer playbooks under agents/ are reference material, not auto-dispatched; see the agent roster.

Minimum Inputs

• paper.tex, paper.typ, or paper.pdf.
• Optional --venue, --lang en|zh, and --report-style deep-review|peer-review.
• Optional --focus full|editor|theory|literature|methodology|logic for deep-review.
• --previous-report for re-audit.
• --review-dir when continuing or rendering an existing review workspace.

Script Entry Points

Script	Purpose
audit.py	Public mode router for quick-audit, deep-review, gate, polish, and re-audit
prepare_review_workspace.py	Creates the review workspace for deep review
build_claim_map.py	Extracts headline claims, closure targets, and claim candidates
check_citations.py / check_references.py	Citation and reference hygiene checks
verify_quotes.py	Verifies report quotes against source text
render_deep_review_report.py	Renders Markdown deep-review reports
render_html_report.py	Renders bilingual HTML report twins
render_revision_trajectory.py	Produces the revision trajectory artifact
diff_review_issues.py	Supports re-audit comparison

Output Artifacts

For deep-review, the workspace root is reader-facing:

• review_report.md and review_report.html
• revision_suggestions.md and revision_suggestions.html

Supporting artifacts live under artifacts/:

• artifacts/summary/paper_summary.md, overall_assessment.txt, peer_review_report.md
• artifacts/data/final_issues.json, all_comments.json, claim_map.json, section_index.json, revision_suggestions.json, revision_trajectory.md
• artifacts/meta/metadata.json, checkpoint.json, phase0_context.md, full_text.md
• artifacts/sections/, artifacts/comments/, artifacts/committee/, and artifacts/references/

The report language is controlled by --lang en|zh. Headings and labels switch language; issue quotes, source tags, and structured field values stay faithful to the source.

Public Resources

References

• Agent Roster
• Audit Guide
• Changelog
• Pre-Submission Checklists
• Claim-Evidence Contract
• Consolidation Rules
• Data Availability Advisory
• Deep Review Criteria
• Editorial Decision Standards
• FORBIDDEN_TERMS.md
• Issue Schema
• Literature Grounding Scoring Guide
• Mode Guide
• Output Layout
• Over-Claim Guard (Audit Lane Reference)
• Polish Guide
• Pre-Submission Mechanical Rules
• PRESUBMISSION Mode Integration
• Qualitative Research Standards Reference
• Quality Rubrics for Paper Audit
• Quick Reference
• Review Criteria
• Review Lane Guide
• Reviewer Psychology
• ScholarEval 9-Dimension Scoring Guide
• Scoring Systems
• Scripts Map
• Reviewer Lane Templates
• Troubleshooting
• Venue-Specific Rules
• Workflow Detail

Templates

• Audit Report Template
• Deep Review Report Template
• Revision Roadmap Template

Examples

• Gate Mode Example Output
• Peer-Review Primary View Example
• Deep Review Example Output
• Self-Check Example Output

Agent Contracts

• Claims vs Evidence Reviewer Agent
• Committee Editor Agent (Pre-Review Screen)
• Committee Reviewer 3 (Literature Dialogue Auditor)
• Committee Reviewer 4 (Logic Chain Auditor)
• Committee Reviewer 2 (Methodology Transparency Inspector)
• Committee Reviewer 1 (Theory Contribution Interrogator)
• Critical Reviewer Agent (Devil's Advocate)
• Domain Reviewer Agent
• Editor-in-Chief Agent (Desk Reject Screener)
• Evaluation Fairness Reviewer Agent
• Literature Reviewer Agent
• Methodology Reviewer Agent
• Notation and Numeric Consistency Reviewer Agent
• Prior-Art Reviewer Agent
• Revision Coach Agent
• Revision Suggestion Agent
• Section Reviewer Agent
• Self-Consistency Reviewer Agent
• Synthesis Agent

Common Requests

Run a quick-audit on paper.tex and tell me what blocks submission.

Deep-review this manuscript like a journal reviewer and produce the review workspace plus HTML report.

Gate this IEEE paper and separate hard blockers from advisory pseudocode recommendations.

Audit only the literature positioning and tell me whether the claimed gap is real or selectively framed.

Re-audit this revised manuscript against report_v1.md and summarize resolved vs unresolved issues.

Important Notes

• PRESUBMISSION is an internal mechanical layer plugged into existing modes, not a separate public mode.
• Full/editor deep-review may promote high-signal pre-submission findings into pre_submission_readiness; other focused reviews keep them as Phase 0 context.
• claim_map.json distinguishes visible anchors from support strength; a citation key alone does not prove support.
• Data availability findings are advisory unless a venue-required central source gap should block submission.
• PDF input runs text-only checks and skips LaTeX/Typst source hygiene.