Paper Audit Skill v4.2
paper-audit is now deep-review-first. Its core job is to behave like a serious reviewer: find technical, methodological, claim-level, and cross-section issues; keep script-backed findings separate from reviewer judgment; and return a structured issue bundle plus a revision roadmap.
Use it for audit and review. Do not use it as the first tool for source editing, sentence rewriting, or build fixing.
What This Skill Produces
quick-audit: fast submission-readiness screen with script-backed findings
deep-review: reviewer-style structured issue bundle with major/moderate/minor findings
gate: PASS/FAIL decision calibrated for submission blockers
re-audit: compare current issue bundle against a previous audit
polish: precheck-only handoff into a polishing workflow
The primary product is no longer just a score. For deep-review, the main outputs are:
final_issues.json
overall_assessment.txt
review_report.md
revision_roadmap.md
Do Not Use
- direct source surgery on
.tex / .typ
- compilation debugging as the main task
- free-form literature survey writing
- cosmetic grammar cleanup without an audit goal
Critical Rules
- Never rewrite the paper source unless the user explicitly switches to an editing skill.
- Never fabricate references, baselines, or reviewer evidence.
- Always distinguish
[Script] from [LLM] findings.
- Always anchor reviewer findings to a quote, section, or exact textual location.
- Be conservative with OCR noise, formatting quirks, and obvious copy-editing trivia.
- Review like a careful reader: understand the author's intended meaning before flagging an issue.
Mode Selection
| Requested intent |
Mode |
| "check my paper", "quick audit", "submission readiness" |
quick-audit |
| "review my paper", "simulate peer review", "harsh review", "deep review" |
deep-review |
| "is this ready to submit", "gate this submission", "blockers only" |
gate |
| "did I fix these issues", "re-audit", "compare against old review" |
re-audit |
| "polish the writing, but only if safe" |
polish |
Legacy aliases still work for one compatibility cycle:
self-check -> quick-audit
review -> deep-review
Committee Focus Routing (deep-review)
For deep-review, use the Academic Pre-Review Committee by default. This is a 5-role review pass:
- Editor (desk-reject screen)
- Reviewer 1 (theory contribution)
- Reviewer 3 (literature dialogue / gap)
- Reviewer 2 (methodology transparency)
- Reviewer 4 (logic chain)
If the user requests a single dimension, run only the matching committee role(s).
If --focus ... is provided, it overrides keyword inference:
--focus full (default)
--focus editor|theory|literature|methodology|logic
Keyword map (English + Chinese):
- editor: "desk reject", "pre-screen", "editor", "EIC", "主编", "预筛", "初筛"
- theory: "theory", "contribution", "novelty", "theoretical dialogue", "理论", "贡献", "创新性"
- literature: "related work", "literature", "research gap", "citation", "文献", "综述", "Research Gap", "引用"
- methodology: "methods", "sample", "coding", "data", "design", "SRQR", "方法", "样本", "编码", "数据", "研究设计", "透明度"
- logic: "logic", "argument", "causal", "structure", "论证", "因果", "逻辑", "结构"
Output language: match the user's request language. If ambiguous, match the paper language.
Review Standard
Read these references before running reviewer-style work:
references/REVIEW_CRITERIA.md
references/DEEP_REVIEW_CRITERIA.md
references/CHECKLIST.md
references/CONSOLIDATION_RULES.md
references/ISSUE_SCHEMA.md
The deep-review workflow uses a 16-part issue taxonomy:
- formula / derivation errors
- notation inconsistency
- prose vs formal object mismatch
- numerical inconsistency
- missing justification
- overclaim or claim inaccuracy
- ambiguity that can mislead a careful reader
- underspecified methods / missing information
- internal contradiction
- self-consistency of standards
- table structure violations
- abstract structural incompleteness
- theory contribution deficiency
- qualitative methodology opacity
- pseudo-innovation / straw man
- paragraph-level argument incoherence
Workflow
Common Step 0
Parse $ARGUMENTS and infer the mode if the user did not provide one. State the inferred mode before running commands if you had to infer it.
quick-audit
- Run:
uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode quick-audit ...
- Present a concise report:
Submission Blockers first
- then
Quality Improvements
- then checklist items
- mark quick-audit findings with
[Script] provenance
- If the user clearly wants reviewer-depth critique after the quick screen, escalate to
deep-review.
deep-review
Use this as the default reviewer-style path.
Phase 1: Prepare workspace
Run:
uv run python -B "$SKILL_DIR/scripts/prepare_review_workspace.py" <paper> --output-dir ./review_results
This creates:
full_text.md
metadata.json
section_index.json
claim_map.json
paper_summary.md
sections/*.md
comments/
references/ (minimal copies for reviewer agents)
committee/ (committee reviewer artifacts)
Phase 2: Phase 0 automated audit
Run:
uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode deep-review ...
Treat this as Phase 0 only. It supplies script-backed context and scores, not the final review.
Phase 3: Committee + Review Lanes
Phase 3A: Academic Pre-Review Committee (default)
Decide committee focus:
- If
--focus ... is provided, use it.
- Otherwise infer from the user request using the keyword map in "Committee Focus Routing".
- If nothing matches, default to
full (all five roles).
Dispatch the committee reviewers (in this exact order) and have them write artifacts into the workspace:
agents/committee_editor_agent.md
- write:
committee/editor.md
- write:
comments/committee_editor.json
agents/committee_theory_agent.md
- write:
committee/theory.md
- write:
comments/committee_theory.json
agents/committee_literature_agent.md
- write:
committee/literature.md
- write:
comments/committee_literature.json
agents/committee_methodology_agent.md
- write:
committee/methodology.md
- write:
comments/committee_methodology.json
agents/committee_logic_agent.md
- write:
committee/logic.md
- write:
comments/committee_logic.json
If subagents are unavailable, run the committee reviewers inline, but keep the same file outputs.
Then write: committee/consensus.md
- include: overall score (1-10), ordered priorities, and the top 3 issues to fix first
- scoring formula:
- start at 9.0
- subtract:
1.5 * (# major) + 0.7 * (# moderate) + 0.2 * (# minor)
- floor at 1.0
- if Editor verdict is Desk Reject, cap at 4.0
Note: render_deep_review_report.py automatically embeds committee/*.md into review_report.md when present.
Phase 3B: Section and cross-cutting review lanes (coverage)
Read:
references/SUBAGENT_TEMPLATES.md
references/REVIEW_LANE_GUIDE.md
Then dispatch reviewer tasks for:
- section lanes
- introduction / related work
- methods
- results
- discussion / conclusion
- appendix, if present
- cross-cutting lanes
- claims vs evidence
- notation and numeric consistency
- evaluation fairness and reproducibility
- self-standard consistency
- prior-art and novelty grounding
Each lane writes a JSON array into comments/.
If subagents are unavailable, use the built-in deterministic fallback lane pass in scripts/audit.py so the workflow still writes lane-compatible JSON into comments/ before consolidation.
Phase 4: Consolidation
Run:
uv run python -B "$SKILL_DIR/scripts/consolidate_review_findings.py" <review_dir>
uv run python -B "$SKILL_DIR/scripts/verify_quotes.py" <review_dir> --write-back
uv run python -B "$SKILL_DIR/scripts/render_deep_review_report.py" <review_dir>
Consolidation rules:
- merge exact duplicates
- keep distinct paper-level consequences separate even if they share a root cause
- preserve singleton findings unless clearly false positive
- assign
comment_type, severity, confidence, and root_cause_key
Phase 5: Present result
Summarize:
- 1 short paragraph overall assessment
- counts of major / moderate / minor issues
- 3 highest-priority revision items
- path to
review_report.md and final_issues.json
gate
- Run:
uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode gate ...
- EIC Screening (Phase 0.5): Read
agents/editor_in_chief_agent.md and perform the editor-in-chief desk-reject screening on the paper's title, abstract, and introduction. This evaluates pitch quality, venue fit, fatal flaws, and presentation baseline. A desk-reject verdict is a gate blocker.
- Report PASS/FAIL.
- Present EIC screening results first (verdict + score + justification).
- List blockers next.
- Keep advisory items separate from blockers.
- For IEEE pseudocode checks, make it explicit which issues are mandatory and which are only IEEE-safe recommendations.
re-audit
- Requires
--previous-report PATH.
- Run:
uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode re-audit --previous-report <path> ...
- If both old and new
final_issues.json bundles are available, also run:uv run python -B "$SKILL_DIR/scripts/diff_review_issues.py" <old_final_issues.json> <new_final_issues.json>
- Present:
- root-cause-aware status labels:
FULLY_ADDRESSED, PARTIALLY_ADDRESSED, NOT_ADDRESSED, NEW
- use structured prior issue bundles when available, but still accept Markdown previous reports
polish
- Run the audit precheck:
uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode polish ...
- If blockers exist, stop and report them.
- Only proceed into polishing if the precheck is safe.
Output Contract
For deep-review, the final issue schema is:
{
"title": "short issue title",
"quote": "exact quote from paper",
"explanation": "why this matters and what remains problematic",
"comment_type": "methodology|claim_accuracy|presentation|missing_information",
"severity": "major|moderate|minor",
"confidence": "high|medium|low",
"source_kind": "script|llm",
"source_section": "methods",
"related_sections": ["results", "appendix"],
"root_cause_key": "shared-normalized-key",
"review_lane": "claims_vs_evidence",
"gate_blocker": false,
"quote_verified": true
}
Always prefer:
- exact quotes over vague paraphrase
- evidence-backed findings over style commentary
- issue bundle + roadmap over raw script dumps
References
| File |
Purpose |
references/REVIEW_CRITERIA.md |
top-level audit scoring and mapping |
references/DEEP_REVIEW_CRITERIA.md |
deep-review-specific issue taxonomy (16 dimensions) and leniency rules |
references/CONSOLIDATION_RULES.md |
deduplication and root-cause merge policy |
references/ISSUE_SCHEMA.md |
canonical JSON schema |
references/REVIEW_LANE_GUIDE.md |
section lanes and cross-cutting lanes |
references/SUBAGENT_TEMPLATES.md |
reviewer task templates |
references/QUICK_REFERENCE.md |
CLI and mode cheat sheet |
Scripts
| Script |
Purpose |
scripts/audit.py |
Phase 0 audit and mode entrypoint |
scripts/prepare_review_workspace.py |
create deep-review workspace |
scripts/build_claim_map.py |
extract headline claims and closure targets |
scripts/consolidate_review_findings.py |
deduplicate comment JSONs |
scripts/verify_quotes.py |
verify exact quote presence |
scripts/render_deep_review_report.py |
render final Markdown report |
scripts/diff_review_issues.py |
compare old vs new issue bundles |
Reviewer Lanes
Committee agents (deep-review default):
committee_editor_agent.md
committee_theory_agent.md
committee_literature_agent.md
committee_methodology_agent.md
committee_logic_agent.md
Default deep-review lanes live in agents/:
section_reviewer_agent.md
claims_evidence_reviewer_agent.md
notation_consistency_reviewer_agent.md
evaluation_fairness_reviewer_agent.md
self_consistency_reviewer_agent.md
prior_art_reviewer_agent.md
synthesis_agent.md
editor_in_chief_agent.md — EIC desk-reject screener (used in gate mode)
Specialized deep-review agents (read their files for activation criteria):
critical_reviewer_agent.md — devil's advocate with C3-C5 checks
domain_reviewer_agent.md — domain expertise with A1-A7 assessments
methodology_reviewer_agent.md — methodology rigor with B3-B10 checks
literature_reviewer_agent.md — evidence-based literature verification (optional, --literature-search)
Examples
- “Review this manuscript like a serious conference reviewer and tell me the biggest validity risks.”
- “Run a quick audit on
paper.tex and tell me what blocks submission.”
- “Gate this IEEE submission and separate blockers from recommendations.”
- “Re-audit this revision against my previous report.”
1---2name: paper-audit3description: Deep-review-first audit for Chinese and English academic papers across LaTeX, Typst, and PDF formats. Use whenever the user wants reviewer-style paper critique, pre-submission readiness checks, pass/fail gate decisions, structured revision roadmaps, or re-audits of revised manuscripts. Trigger even if the user only says "review my paper", "check if this is ready to submit", "audit this PDF", "simulate peer review", "find the biggest problems in this manuscript", or "re-check whether I fixed the review issues". Do not use for direct source editing or compilation-heavy repair; route those to the format-specific writing skills instead.4metadata:5 source: https://skillmd.com/skills/qhjqhj00/paper-audit6 category: academic-writing7 tags: [audit, deep-review, paper, pdf, latex, typst, chinese, english, reviewer, gate, re-audit]8 version: "4.2"9 last_updated: "2026-04-02"10argument-hint: "[paper.tex|paper.typ|paper.pdf] [--mode quick-audit|deep-review|gate|re-audit|polish] [--focus full|editor|theory|literature|methodology|logic] [--venue VENUE] [--previous-report PATH] [--literature-search] [--scholar-eval] [--format markdown|json]"11allowed-tools: Read, Glob, Grep, Bash(uv *), Task12---1314# Paper Audit Skill v4.21516`paper-audit` is now **deep-review-first**. Its core job is to behave like a serious reviewer: find technical, methodological, claim-level, and cross-section issues; keep script-backed findings separate from reviewer judgment; and return a structured issue bundle plus a revision roadmap.1718Use it for audit and review. Do not use it as the first tool for source editing, sentence rewriting, or build fixing.1920## What This Skill Produces2122- `quick-audit`: fast submission-readiness screen with script-backed findings23- `deep-review`: reviewer-style structured issue bundle with major/moderate/minor findings24- `gate`: PASS/FAIL decision calibrated for submission blockers25- `re-audit`: compare current issue bundle against a previous audit26- `polish`: precheck-only handoff into a polishing workflow2728The primary product is no longer just a score. For `deep-review`, the main outputs are:2930- `final_issues.json`31- `overall_assessment.txt`32- `review_report.md`33- `revision_roadmap.md`3435## Do Not Use3637- direct source surgery on `.tex` / `.typ`38- compilation debugging as the main task39- free-form literature survey writing40- cosmetic grammar cleanup without an audit goal4142## Critical Rules4344- Never rewrite the paper source unless the user explicitly switches to an editing skill.45- Never fabricate references, baselines, or reviewer evidence.46- Always distinguish `[Script]` from `[LLM]` findings.47- Always anchor reviewer findings to a quote, section, or exact textual location.48- Be conservative with OCR noise, formatting quirks, and obvious copy-editing trivia.49- Review like a careful reader: understand the author's intended meaning before flagging an issue.5051## Mode Selection5253| Requested intent | Mode |54|---|---|55| "check my paper", "quick audit", "submission readiness" | `quick-audit` |56| "review my paper", "simulate peer review", "harsh review", "deep review" | `deep-review` |57| "is this ready to submit", "gate this submission", "blockers only" | `gate` |58| "did I fix these issues", "re-audit", "compare against old review" | `re-audit` |59| "polish the writing, but only if safe" | `polish` |6061Legacy aliases still work for one compatibility cycle:6263- `self-check` -> `quick-audit`64- `review` -> `deep-review`6566## Committee Focus Routing (deep-review)6768For `deep-review`, use the **Academic Pre-Review Committee** by default. This is a 5-role review pass:69701. Editor (desk-reject screen)712. Reviewer 1 (theory contribution)723. Reviewer 3 (literature dialogue / gap)734. Reviewer 2 (methodology transparency)745. Reviewer 4 (logic chain)7576If the user requests a single dimension, run only the matching committee role(s).7778If `--focus ...` is provided, it overrides keyword inference:79- `--focus full` (default)80- `--focus editor|theory|literature|methodology|logic`8182Keyword map (English + Chinese):83- editor: "desk reject", "pre-screen", "editor", "EIC", "主编", "预筛", "初筛"84- theory: "theory", "contribution", "novelty", "theoretical dialogue", "理论", "贡献", "创新性"85- literature: "related work", "literature", "research gap", "citation", "文献", "综述", "Research Gap", "引用"86- methodology: "methods", "sample", "coding", "data", "design", "SRQR", "方法", "样本", "编码", "数据", "研究设计", "透明度"87- logic: "logic", "argument", "causal", "structure", "论证", "因果", "逻辑", "结构"8889Output language: match the user's request language. If ambiguous, match the paper language.9091## Review Standard9293Read these references before running reviewer-style work:94951. `references/REVIEW_CRITERIA.md`962. `references/DEEP_REVIEW_CRITERIA.md`973. `references/CHECKLIST.md`984. `references/CONSOLIDATION_RULES.md`995. `references/ISSUE_SCHEMA.md`100101The deep-review workflow uses a 16-part issue taxonomy:1021031. formula / derivation errors1042. notation inconsistency1053. prose vs formal object mismatch1064. numerical inconsistency1075. missing justification1086. overclaim or claim inaccuracy1097. ambiguity that can mislead a careful reader1108. underspecified methods / missing information1119. internal contradiction11210. self-consistency of standards11311. table structure violations11412. abstract structural incompleteness11513. theory contribution deficiency11614. qualitative methodology opacity11715. pseudo-innovation / straw man11816. paragraph-level argument incoherence119120## Workflow121122### Common Step 0123124Parse `$ARGUMENTS` and infer the mode if the user did not provide one. State the inferred mode before running commands if you had to infer it.125126### `quick-audit`1271281. Run:129 ```bash130 uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode quick-audit ...131 ```1322. Present a concise report:133 - `Submission Blockers` first134 - then `Quality Improvements`135 - then checklist items136 - mark quick-audit findings with `[Script]` provenance1373. If the user clearly wants reviewer-depth critique after the quick screen, escalate to `deep-review`.138139### `deep-review`140141Use this as the default reviewer-style path.142143#### Phase 1: Prepare workspace144145Run:146147```bash148uv run python -B "$SKILL_DIR/scripts/prepare_review_workspace.py" <paper> --output-dir ./review_results149```150151This creates:152153- `full_text.md`154- `metadata.json`155- `section_index.json`156- `claim_map.json`157- `paper_summary.md`158- `sections/*.md`159- `comments/`160- `references/` (minimal copies for reviewer agents)161- `committee/` (committee reviewer artifacts)162163#### Phase 2: Phase 0 automated audit164165Run:166167```bash168uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode deep-review ...169```170171Treat this as **Phase 0 only**. It supplies script-backed context and scores, not the final review.172173#### Phase 3: Committee + Review Lanes174175##### Phase 3A: Academic Pre-Review Committee (default)176177Decide committee focus:178- If `--focus ...` is provided, use it.179- Otherwise infer from the user request using the keyword map in "Committee Focus Routing".180- If nothing matches, default to `full` (all five roles).181182Dispatch the committee reviewers (in this exact order) and have them write artifacts into the workspace:1831841. `agents/committee_editor_agent.md`185 - write: `committee/editor.md`186 - write: `comments/committee_editor.json`1872. `agents/committee_theory_agent.md`188 - write: `committee/theory.md`189 - write: `comments/committee_theory.json`1903. `agents/committee_literature_agent.md`191 - write: `committee/literature.md`192 - write: `comments/committee_literature.json`1934. `agents/committee_methodology_agent.md`194 - write: `committee/methodology.md`195 - write: `comments/committee_methodology.json`1965. `agents/committee_logic_agent.md`197 - write: `committee/logic.md`198 - write: `comments/committee_logic.json`199200If subagents are unavailable, run the committee reviewers inline, but keep the same file outputs.201202Then write: `committee/consensus.md`203- include: overall score (1-10), ordered priorities, and the top 3 issues to fix first204- scoring formula:205 - start at 9.0206 - subtract: `1.5 * (# major) + 0.7 * (# moderate) + 0.2 * (# minor)`207 - floor at 1.0208 - if Editor verdict is Desk Reject, cap at 4.0209210Note: `render_deep_review_report.py` automatically embeds `committee/*.md` into `review_report.md` when present.211212##### Phase 3B: Section and cross-cutting review lanes (coverage)213214Read:215216- `references/SUBAGENT_TEMPLATES.md`217- `references/REVIEW_LANE_GUIDE.md`218219Then dispatch reviewer tasks for:220221- section lanes222 - introduction / related work223 - methods224 - results225 - discussion / conclusion226 - appendix, if present227- cross-cutting lanes228 - claims vs evidence229 - notation and numeric consistency230 - evaluation fairness and reproducibility231 - self-standard consistency232 - prior-art and novelty grounding233234Each lane writes a JSON array into `comments/`.235236If subagents are unavailable, use the built-in deterministic fallback lane pass in `scripts/audit.py` so the workflow still writes lane-compatible JSON into `comments/` before consolidation.237238#### Phase 4: Consolidation239240Run:241242```bash243uv run python -B "$SKILL_DIR/scripts/consolidate_review_findings.py" <review_dir>244uv run python -B "$SKILL_DIR/scripts/verify_quotes.py" <review_dir> --write-back245uv run python -B "$SKILL_DIR/scripts/render_deep_review_report.py" <review_dir>246```247248Consolidation rules:249250- merge exact duplicates251- keep distinct paper-level consequences separate even if they share a root cause252- preserve singleton findings unless clearly false positive253- assign `comment_type`, `severity`, `confidence`, and `root_cause_key`254255#### Phase 5: Present result256257Summarize:258259- 1 short paragraph overall assessment260- counts of major / moderate / minor issues261- 3 highest-priority revision items262- path to `review_report.md` and `final_issues.json`263264### `gate`2652661. Run:267 ```bash268 uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode gate ...269 ```2702. **EIC Screening** (Phase 0.5): Read `agents/editor_in_chief_agent.md` and perform the editor-in-chief desk-reject screening on the paper's title, abstract, and introduction. This evaluates pitch quality, venue fit, fatal flaws, and presentation baseline. A desk-reject verdict is a gate blocker.2713. Report PASS/FAIL.2724. Present EIC screening results first (verdict + score + justification).2735. List blockers next.2746. Keep advisory items separate from blockers.2757. For IEEE pseudocode checks, make it explicit which issues are mandatory and which are only IEEE-safe recommendations.276277### `re-audit`2782791. Requires `--previous-report PATH`.2802. Run:281 ```bash282 uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode re-audit --previous-report <path> ...283 ```2843. If both old and new `final_issues.json` bundles are available, also run:285 ```bash286 uv run python -B "$SKILL_DIR/scripts/diff_review_issues.py" <old_final_issues.json> <new_final_issues.json>287 ```2884. Present:289 - root-cause-aware status labels: `FULLY_ADDRESSED`, `PARTIALLY_ADDRESSED`, `NOT_ADDRESSED`, `NEW`290 - use structured prior issue bundles when available, but still accept Markdown previous reports291292### `polish`2932941. Run the audit precheck:295 ```bash296 uv run python -B "$SKILL_DIR/scripts/audit.py" <paper> --mode polish ...297 ```2982. If blockers exist, stop and report them.2993. Only proceed into polishing if the precheck is safe.300301## Output Contract302303For `deep-review`, the final issue schema is:304305```json306{307 "title": "short issue title",308 "quote": "exact quote from paper",309 "explanation": "why this matters and what remains problematic",310 "comment_type": "methodology|claim_accuracy|presentation|missing_information",311 "severity": "major|moderate|minor",312 "confidence": "high|medium|low",313 "source_kind": "script|llm",314 "source_section": "methods",315 "related_sections": ["results", "appendix"],316 "root_cause_key": "shared-normalized-key",317 "review_lane": "claims_vs_evidence",318 "gate_blocker": false,319 "quote_verified": true320}321```322323Always prefer:324325- exact quotes over vague paraphrase326- evidence-backed findings over style commentary327- issue bundle + roadmap over raw script dumps328329## References330331| File | Purpose |332|---|---|333| `references/REVIEW_CRITERIA.md` | top-level audit scoring and mapping |334| `references/DEEP_REVIEW_CRITERIA.md` | deep-review-specific issue taxonomy (16 dimensions) and leniency rules |335| `references/CONSOLIDATION_RULES.md` | deduplication and root-cause merge policy |336| `references/ISSUE_SCHEMA.md` | canonical JSON schema |337| `references/REVIEW_LANE_GUIDE.md` | section lanes and cross-cutting lanes |338| `references/SUBAGENT_TEMPLATES.md` | reviewer task templates |339| `references/QUICK_REFERENCE.md` | CLI and mode cheat sheet |340341## Scripts342343| Script | Purpose |344|---|---|345| `scripts/audit.py` | Phase 0 audit and mode entrypoint |346| `scripts/prepare_review_workspace.py` | create deep-review workspace |347| `scripts/build_claim_map.py` | extract headline claims and closure targets |348| `scripts/consolidate_review_findings.py` | deduplicate comment JSONs |349| `scripts/verify_quotes.py` | verify exact quote presence |350| `scripts/render_deep_review_report.py` | render final Markdown report |351| `scripts/diff_review_issues.py` | compare old vs new issue bundles |352353## Reviewer Lanes354355Committee agents (deep-review default):356357- `committee_editor_agent.md`358- `committee_theory_agent.md`359- `committee_literature_agent.md`360- `committee_methodology_agent.md`361- `committee_logic_agent.md`362363Default deep-review lanes live in `agents/`:364365- `section_reviewer_agent.md`366- `claims_evidence_reviewer_agent.md`367- `notation_consistency_reviewer_agent.md`368- `evaluation_fairness_reviewer_agent.md`369- `self_consistency_reviewer_agent.md`370- `prior_art_reviewer_agent.md`371- `synthesis_agent.md`372- `editor_in_chief_agent.md` — EIC desk-reject screener (used in `gate` mode)373374Specialized deep-review agents (read their files for activation criteria):375376- `critical_reviewer_agent.md` — devil's advocate with C3-C5 checks377- `domain_reviewer_agent.md` — domain expertise with A1-A7 assessments378- `methodology_reviewer_agent.md` — methodology rigor with B3-B10 checks379- `literature_reviewer_agent.md` — evidence-based literature verification (optional, `--literature-search`)380381## Examples382383- “Review this manuscript like a serious conference reviewer and tell me the biggest validity risks.”384- “Run a quick audit on `paper.tex` and tell me what blocks submission.”385- “Gate this IEEE submission and separate blockers from recommendations.”386- “Re-audit this revision against my previous report.”