PaperJury (CS-conference paper review and editing)
PaperJury edits and hardens any CS-conference paper. It runs in
three modes. In direct-edit mode (the common case) the user describes a change
in Chinese or English and the LaTeX is edited directly through a CS-venue writing
toolkit, with author sign-off. In review mode (occasional, pre-submission) it
exposes the manuscript to a harsh, multi-perspective courtroom review engine that
adjudicates each issue (N holistic domain reviewers -> contestability routing ->
two-sided trial -> three-way verdict, with a polish track and a clerk-converged
multi-round loop), gates every change behind consensus, and tracks issues in a durable
ledger. In auto mode (unattended, opt-in via /goal) it runs that same engine
toward a verifiable goal, applying safe fixes under a drift-bounded policy and
queueing the risky ones for one human pass on return. All modes share the same
writing toolkit, hard rules, ledger, and author sign-off (auto via up-front policy
sign-off plus the queue, see hard rule 1).
This skill is fully generic. It ships no hardcoded paths, no project files,
and no embedded paper. Everything specific to a given paper (where the
manuscript is, the venue, who signs off, the house style) is resolved at runtime
or supplied by a config the project owns. The skill itself is the backbone;
any concrete paper is just an instantiation of it.
Scope: CS conferences only. Three venue families, each with its own style profile:
- Vision: CVPR, ICCV, ECCV, WACV
- NLP: ACL, EMNLP, NAACL, COLING
- ML: ICLR, NeurIPS, ICML, AAAI, COLM
Codex runtime
This repository is the Codex-first PaperJury port. Read codex/runtime.md
before running review or auto, then use codex/phase-contracts.md for phase
inputs, outputs, isolation, and validation. Run deterministic guards with
node scripts/*.js, and run semantic fan-out via Codex subagents only when the
user has explicitly authorized multi-agent work.
Soft update reminder: at the start of each PaperJury invocation, before choosing
the mode or editing a manuscript, run node scripts/check-update.js from the
skill root unless PAPERJURY_DISABLE_UPDATE_CHECK=1 is set. If it reports an
available update, show the notice once and continue. If the check is skipped,
silent, or cannot reach GitHub, continue without mentioning it; update checks are
never allowed to block review or editing.
If host capabilities conflict with this file, keep the scientific invariants:
author sign-off, reviewer isolation, no silent drops, a machine ledger, and honest
degradation when a check cannot be verified.
When to use / when not
Three modes, one skill. Pick by what the user is asking for:
- Direct-edit mode (the common case). The user describes a change in Chinese
(or English) and wants the LaTeX edited directly: "把这段改成...", "polish this
paragraph", "把我对 intro 的想法写成 LaTeX", "tighten this". No review panel; go
straight to drafting the patch through the writing toolkit, with author sign-off.
- Review mode (occasional, pre-submission). The user wants the paper critiqued
or hardened: review / critique / 审稿 / 评审 / mock-review, or iterating a draft
to clear reviewer-raised issues. This runs the courtroom review engine
(
references/review-engine-v3.md).
- Auto mode (unattended). The user opts in via
/goal (or config mode: auto)
to run the review-revise loop AFK toward a verifiable goal. Establish the spine
up front (the one human step), then the engine applies safe fixes under the
bounded-aggressive policy and queues the rest. See references/auto-mode.md.
Never self-detect auto; it is explicit only.
Do NOT use for: writing a paper from scratch (use ml-paper-writing), figure or
diagram generation (use academic-plotting), or an official-venue rebuttal (this
is a pre-submission self-hardening loop, no score gate).
The three primitives
This paradigm is expressed as Skill + Semantic Fan-Out + Memory. Each carries
one concern; together they replace the heavy per-round file-and-flag machinery a
hand-rolled version accumulates.
- Skill (this folder) = entry point + methodology. The protocol, the
reviewer panel, the contestability routing, the writing toolkit, the human gates.
Detail in
references/review-engine-v3.md, references/reviewer-personas.md,
references/writing-toolkit.md.
- Semantic fan-out = reviewer, jury, merge, audit, and drafting agents. In
Codex it is implemented by the orchestrator spawning isolated subagents, or by
an explicitly labeled degraded single-agent path when subagents are unavailable.
The v3 courtroom engine is
assign-reviewers -> reading-check -> coverage-auditor -> merge ->
{trial (+ escalate) || polish} -> recall-audit -> drafter ->
{edit-audit | meaning-audit} -> clerk. Protocol + every orchestrator seam:
references/review-engine-v3.md; Codex runtime mapping:
codex/runtime.md; Codex-native phase contracts:
codex/phase-contracts.md.
- Memory = durable state + learned conventions. Two layers:
- Ledger (
LEDGER.json resolved at runtime = the machine source of truth,
plus a rendered LEDGER.md view; managed by scripts/ledger.js): the live,
mutable issue state across rounds and sessions. Schema + status state machine:
references/ledger-schema.md.
- Project conventions: stable house-style and venue conventions stored in
the active project, for example
.paper-review/CONVENTIONS.md or the host's
native project memory. Do not store transient issue state outside the ledger.
Resolving inputs at runtime (no hardcoded paths)
The skill ships ZERO hardcoded paths or project files. On trigger it resolves
each input by discovery first, then asking:
- manuscript: detect the main source (the
.tex with \documentclass /
\begin{document}, or the file the user names). If several candidates, ask.
- venue_family: the user can name it, or an agent reads the class file to
GUESS the family (e.g. a cvpr/iccv style, an acl style, a neurips/iclr style).
There is no hardcoded venue list and no deterministic detector; if unclear, ask.
- ledger: default to
<manuscript-dir>/.paper-review/LEDGER.json (the machine
source of truth; scripts/ledger.js also renders a LEDGER.md view). Create if
absent, reuse if present. The user may point elsewhere.
- author: ask who signs off on edits (default: the current user). Every edit
needs explicit authorization.
- personas: default to N domain-expert holistic reviewers assigned at runtime
(
assign-reviewers, from the project gatekeeper core + a generated domain overlay);
the three generic lenses in references/reviewer-personas.md are the degrade
fallback. If the project defines its own named reviewer subagents, use them as
agentType; otherwise inline the persona prompts.
- style_profile: start from the venue-family default; refine from any
conventions recalled from memory or pinned in a project config.
A project MAY pin these by dropping a config in ITS OWN repo (see
configs/config-template.md for the shape). That file is owned by the project,
never by this skill. At round start, recall any pinned conventions from memory.
Direct-edit mode (the common case)
The user states a change in Chinese or English; you draft and apply the LaTeX edit.
No panel, no ledger, no discussion. Minimal flow:
- Locate. Resolve the manuscript and find the target passage the instruction
refers to (a paragraph, sentence, caption, table cell). If it is ambiguous on a
large file, ask which passage; do not guess.
- Draft. Pick the writing-toolkit prompt matching the instruction
(
translate-to-english for a Chinese idea, polish-english / de-ai for a
rewrite, compress / expand for length, caption / experiment-analysis
for those units) and draft the LaTeX patch to do exactly what was asked. The
Common guards apply (LaTeX-safe, plain CS prose, no log leakage into the .tex).
- Self-gate. Run
logic-check on the drafted passage.
- Sign-off. Show the patch and get explicit author approval (hard rule 1).
- Apply. Write only the patch into the manuscript; keep any back-translation
or note author-side.
This is the writing toolkit used on its own. Escalate to review mode only when the
user wants the paper critiqued or hardened, not for a single asked-for edit.
Why fan-out is separate from conversation
The reviewer panel and the trial jury are pure fan-out: spawn, collect,
merge. Codex uses subagents when the user has explicitly authorized parallel
agent work. Each reviewer or juror gets only the quoted manuscript context and a
strict JSON contract; no peer report, ledger, prior round, or project file path is
included in the prompt.
But the loop has genuine human gates (the author reviews the issue list, gives
per-issue direction, authorizes edits, breaks ties). Fan-out phases run to
completion and return a result; they do not pause mid-run for hours of human
input. So:
- fan-out steps (reviewers, trial, polish, recall, merge) -> Codex subagents
- human gates (per-issue direction, authorization, override) -> main conversation turns
- cross-round truth (the ledger) + stable conventions -> project-owned memory files or host memory
Review mode: one round, end to end
The full adversarial loop (the v3 courtroom engine). Use it to harden the paper,
not for a single asked-for edit (that is direct-edit mode). Full protocol + the
14 orchestrator seams: references/review-engine-v3.md. [SF] = semantic
fan-out step using Codex subagents per codex/runtime.md and
codex/phase-contracts.md; [det] = deterministic Node guard run
orchestrator-side between semantic steps; [HUMAN] = author gate; [LEDGER] =
state write.
- Resolve + recall. Resolve the inputs above; recall this paper's conventions
from memory. Pick scope:
full (whole paper) or passage (one section / para / claim).
[det] decompose. Split the manuscript into reading units + stable
passage_ids + the canonical section list.
[SF] assign-reviewers + [HUMAN] confirm. Name N subfields (2-4,
default 3); instantiate N holistic domain reviewers from the gatekeeper core + a
generated overlay. An unconfirmable slot degrades per slot to a generic gatekeeper
(the three generic lenses in reviewer-personas.md are the fallback). The author
confirms the assignment (or pins it via config).
[SF] reading-check. Each reviewer reads the WHOLE paper → weaknesses
{significance(major|minor), kind(mechanical|substantive), verbatim quote —
cannot quote = did not read} + one overall_confidence + a per-section coverage
report. Anti-skim is three layers: [det] per-section quote-verify, [SF]
coverage-auditor, [SF] targeted re-invoke.
[SF] merge. Semantic dedup across reviewers; derive significance (MAX) /
kind (substantive-dominates) / corroboration. [LEDGER] intake as raised.
[det] route. mechanical → polish; substantive&minor → polish;
substantive&major → trial (two parallel tracks).
[SF] trial. Per substantive-major charge: a whole-paper DEFENSE → 5
decorrelated local-context jurors (+ on-demand expansion) → a deterministic verdict
(decide iff quorum surviving >= ceil(0.8*jurySize) AND one side > 60% of
surviving votes; else escalate to 12). Verdict ∈ {invalid-drop, valid-fixable,
author-required, escalate}; the judge sets a close_criterion ONLY for a
valid-fixable charge, satisfiable by editing existing text (no new data). [SF]
polish runs the off-gate mechanical/minor track in parallel (never silently dropped).
[SF] recall-audit. Mode A revives wrongly-dropped charges; Mode B spot-checks
strong-consensus majors BEFORE the edit. Runs before the drafter.
[HUMAN] Authorize + [SF] drafter + edit-safety. On authorization, the
drafter writes the minimal patch per surviving valid-fixable. The edit-safety chain
gates it: [det] anchor-diff + cross-ref → [SF] meaning-audit (frozen anchor,
four-state) / edit-audit (risky non-anchor); [det] apply-patch + compile-guard land
a passing patch and [LEDGER] mark closed; a drift / anchor / failed edit is
reverted and queued. Revision logs / back-translations stay author-side.
[SF] clerk + report. The clerk reconciles the round boundary (carried
open-questions vs this round's edits, via a passage_id + similarity merge key) and
emits convergence counts. Summarize new/closed counts; in review mode do not
auto-start the next round (auto mode drives the outer loop via /goal).
GATE: node scripts/ledger.js gate = 0 gate-blocking active major (gate-blocking =
{raised, in-trial, re-trial, valid-fixable}; author-required / queued / dropped /
closed are gate-OK and author-required accumulates to the queue). Full protocol +
ledger schema + status machine: references/review-engine-v3.md,
references/ledger-schema.md. The single-pass 3-reviewer panel is available as
the review-panel fast path in codex/phase-contracts.md.
Hard rules (load-bearing, venue-agnostic)
- Never edit the manuscript without explicit author sign-off. Auto-mode
carve-out: the rule HOLDS; auto satisfies it via UP-FRONT sign-off (the spine
confirmation + the pre-authorized bounded-aggressive policy) plus the return
queue, not per-edit sign-off. Nothing outside the authorized envelope is applied.
- Reviewers / jurors are isolated. Fresh eyes per round: no cross-talk, no
prior-round leakage, no sight of the ledger. Enforced by (a) what goes into each
agent's prompt AND (b) an explicit ISOLATION instruction in every reviewer-type
prompt telling the agent to judge only the quoted text and not read files.
- A valid-fixable issue carries a
close_criterion (one concrete sentence an
edit must satisfy), set by the judge at trial; it is null at intake.
- No leakage into the reviewed text. Revision logs, back-translations, and
self-check verdicts are author-side aids; they never enter the manuscript or
any frozen snapshot.
- Disagreement resolves through discussion, then override (logged), never a
silent dismissal.
- No hardcoded paths or project files in the skill. Resolve at runtime.
Memory convention
- At round start: recall the paper's conventions (house style, venue, persona
tuning) from memory; read the resolved
LEDGER.json for open issues.
- During the round: the ledger is the only mutable truth; update it at merge,
trial verdicts, recall, and close.
- After the round: persist any newly learned stable convention to memory (e.g. a
house-style rule a reviewer surfaced), not the transient issue state.
Intensity and host-specific launch notes
The quick panel can run in stronger or cheaper forms using the review-panel
fast path in codex/phase-contracts.md:
- loop-until-dry: re-runs independent fresh panels and accumulates only issues
not seen before, stopping after
dryStop consecutive passes that add no
surviving issue (hard cap maxRounds). Raises recall past a single pass.
- adversarial verify: each new issue faces perspective-diverse skeptics
(misreading / already-addressed / scope-or-severity) and is kept unless a
majority refute it, filtering plausible-but-wrong issues before they reach the
ledger. Bias is to keep, so real flaws are not lost.
Toggle via args: thorough mode -> defaults (maxRounds 4, dryStop 2,
verify true); light mode -> pass {maxRounds:1, verify:false} for the basic
single-panel form. The loop is budget-aware and stops early if the token budget
runs low. Codex treats intensity as an ordinary runtime setting, not as a
separate launch keyword.
Capabilities and status
Built: the review engine; the submission-readiness checker (deterministic desk-reject screening plus a real LaTeX compile, degrading to a structural lint when no toolchain is present); and auto mode (the review-revise loop toward a goal under a drift-bounded policy, applying safe fixes and queueing risky ones for author review). Roadmap: vision-based layout verification, automatic venue detection from the class file, and reviewer personas tuned to each venue community.
Related skills
ml-paper-writing: from-scratch drafting, citation verification (never
hallucinate citations), conference checklists. This loop borrows its
sentence-level guidance for the edit-drafting step rather than duplicating it.
academic-plotting: figure and architecture-diagram generation (out of scope
here; this loop edits text and captions, not figure images).
1---2name: paperjury-33description: Pre-submission CS-conference LaTeX paper editing and adversarial review. Use for direct edits such as polish, de-AI, translating Chinese notes to LaTeX, compressing passages, captions, and experiment prose; for review or critique requests such as review, mock-review, 审稿, 评审; and for explicit auto or goal-driven review-revise loops. Runs deterministic Node guards, a durable ledger, isolated reviewer and juror fan-out, author sign-off, and compile or compliance checks. Not for writing a paper from scratch, figure generation, or official rebuttals.4---56# PaperJury (CS-conference paper review and editing)78PaperJury edits and hardens any CS-conference paper. It runs in9three modes. In **direct-edit mode** (the common case) the user describes a change10in Chinese or English and the LaTeX is edited directly through a CS-venue writing11toolkit, with author sign-off. In **review mode** (occasional, pre-submission) it12exposes the manuscript to a harsh, multi-perspective courtroom review engine that13adjudicates each issue (N holistic domain reviewers -> contestability routing ->14two-sided trial -> three-way verdict, with a polish track and a clerk-converged15multi-round loop), gates every change behind consensus, and tracks issues in a durable16ledger. In **auto mode** (unattended, opt-in via `/goal`) it runs that same engine17toward a verifiable goal, applying safe fixes under a drift-bounded policy and18queueing the risky ones for one human pass on return. All modes share the same19writing toolkit, hard rules, ledger, and author sign-off (auto via up-front policy20sign-off plus the queue, see hard rule 1).2122This skill is **fully generic**. It ships no hardcoded paths, no project files,23and no embedded paper. Everything specific to a given paper (where the24manuscript is, the venue, who signs off, the house style) is resolved at runtime25or supplied by a config the *project* owns. The skill itself is the backbone;26any concrete paper is just an instantiation of it.2728Scope: CS conferences only. Three venue families, each with its own style profile:29- **Vision**: CVPR, ICCV, ECCV, WACV30- **NLP**: ACL, EMNLP, NAACL, COLING31- **ML**: ICLR, NeurIPS, ICML, AAAI, COLM3233## Codex runtime3435This repository is the Codex-first PaperJury port. Read `codex/runtime.md`36before running `review` or `auto`, then use `codex/phase-contracts.md` for phase37inputs, outputs, isolation, and validation. Run deterministic guards with38`node scripts/*.js`, and run semantic fan-out via Codex subagents only when the39user has explicitly authorized multi-agent work.4041Soft update reminder: at the start of each PaperJury invocation, before choosing42the mode or editing a manuscript, run `node scripts/check-update.js` from the43skill root unless `PAPERJURY_DISABLE_UPDATE_CHECK=1` is set. If it reports an44available update, show the notice once and continue. If the check is skipped,45silent, or cannot reach GitHub, continue without mentioning it; update checks are46never allowed to block review or editing.4748If host capabilities conflict with this file, keep the scientific invariants:49author sign-off, reviewer isolation, no silent drops, a machine ledger, and honest50degradation when a check cannot be verified.5152## When to use / when not5354Three modes, one skill. Pick by what the user is asking for:55- **Direct-edit mode (the common case).** The user describes a change in Chinese56 (or English) and wants the LaTeX edited directly: "把这段改成...", "polish this57 paragraph", "把我对 intro 的想法写成 LaTeX", "tighten this". No review panel; go58 straight to drafting the patch through the writing toolkit, with author sign-off.59- **Review mode (occasional, pre-submission).** The user wants the paper critiqued60 or hardened: review / critique / 审稿 / 评审 / mock-review, or iterating a draft61 to clear reviewer-raised issues. This runs the courtroom review engine62 (`references/review-engine-v3.md`).63- **Auto mode (unattended).** The user opts in via `/goal` (or config `mode: auto`)64 to run the review-revise loop AFK toward a verifiable goal. Establish the spine65 up front (the one human step), then the engine applies safe fixes under the66 bounded-aggressive policy and queues the rest. See `references/auto-mode.md`.67 Never self-detect auto; it is explicit only.6869Do NOT use for: writing a paper from scratch (use `ml-paper-writing`), figure or70diagram generation (use `academic-plotting`), or an official-venue rebuttal (this71is a pre-submission self-hardening loop, no score gate).7273## The three primitives7475This paradigm is expressed as **Skill + Semantic Fan-Out + Memory**. Each carries76one concern; together they replace the heavy per-round file-and-flag machinery a77hand-rolled version accumulates.78791. **Skill (this folder) = entry point + methodology.** The protocol, the80 reviewer panel, the contestability routing, the writing toolkit, the human gates.81 Detail in `references/review-engine-v3.md`, `references/reviewer-personas.md`,82 `references/writing-toolkit.md`.832. **Semantic fan-out = reviewer, jury, merge, audit, and drafting agents.** In84 Codex it is implemented by the orchestrator spawning isolated subagents, or by85 an explicitly labeled degraded single-agent path when subagents are unavailable.86 The v3 courtroom engine is87 `assign-reviewers` -> `reading-check` -> `coverage-auditor` -> `merge` ->88 {`trial` (+ escalate) || `polish`} -> `recall-audit` -> `drafter` ->89 {`edit-audit` | `meaning-audit`} -> `clerk`. Protocol + every orchestrator seam:90 `references/review-engine-v3.md`; Codex runtime mapping:91 `codex/runtime.md`; Codex-native phase contracts:92 `codex/phase-contracts.md`.933. **Memory = durable state + learned conventions.** Two layers:94 - **Ledger** (`LEDGER.json` resolved at runtime = the machine source of truth,95 plus a rendered `LEDGER.md` view; managed by `scripts/ledger.js`): the live,96 mutable issue state across rounds and sessions. Schema + status state machine:97 `references/ledger-schema.md`.98 - **Project conventions**: stable house-style and venue conventions stored in99 the active project, for example `.paper-review/CONVENTIONS.md` or the host's100 native project memory. Do not store transient issue state outside the ledger.101102## Resolving inputs at runtime (no hardcoded paths)103104The skill ships ZERO hardcoded paths or project files. On trigger it resolves105each input by **discovery first, then asking**:106107- **manuscript**: detect the main source (the `.tex` with `\documentclass` /108 `\begin{document}`, or the file the user names). If several candidates, ask.109- **venue_family**: the user can name it, or an agent reads the class file to110 GUESS the family (e.g. a cvpr/iccv style, an acl style, a neurips/iclr style).111 There is no hardcoded venue list and no deterministic detector; if unclear, ask.112- **ledger**: default to `<manuscript-dir>/.paper-review/LEDGER.json` (the machine113 source of truth; `scripts/ledger.js` also renders a `LEDGER.md` view). Create if114 absent, reuse if present. The user may point elsewhere.115- **author**: ask who signs off on edits (default: the current user). Every edit116 needs explicit authorization.117- **personas**: default to N domain-expert holistic reviewers assigned at runtime118 (`assign-reviewers`, from the project gatekeeper core + a generated domain overlay);119 the three generic lenses in `references/reviewer-personas.md` are the degrade120 fallback. If the project defines its own named reviewer subagents, use them as121 `agentType`; otherwise inline the persona prompts.122- **style_profile**: start from the venue-family default; refine from any123 conventions recalled from memory or pinned in a project config.124125A project MAY pin these by dropping a config in ITS OWN repo (see126`configs/config-template.md` for the shape). That file is owned by the project,127never by this skill. At round start, recall any pinned conventions from memory.128129## Direct-edit mode (the common case)130131The user states a change in Chinese or English; you draft and apply the LaTeX edit.132No panel, no ledger, no discussion. Minimal flow:1331341. **Locate.** Resolve the manuscript and find the target passage the instruction135 refers to (a paragraph, sentence, caption, table cell). If it is ambiguous on a136 large file, ask which passage; do not guess.1372. **Draft.** Pick the writing-toolkit prompt matching the instruction138 (`translate-to-english` for a Chinese idea, `polish-english` / `de-ai` for a139 rewrite, `compress` / `expand` for length, `caption` / `experiment-analysis`140 for those units) and draft the LaTeX patch to do exactly what was asked. The141 Common guards apply (LaTeX-safe, plain CS prose, no log leakage into the .tex).1423. **Self-gate.** Run `logic-check` on the drafted passage.1434. **Sign-off.** Show the patch and get explicit author approval (hard rule 1).1445. **Apply.** Write only the patch into the manuscript; keep any back-translation145 or note author-side.146147This is the writing toolkit used on its own. Escalate to review mode only when the148user wants the paper critiqued or hardened, not for a single asked-for edit.149150## Why fan-out is separate from conversation151152The reviewer panel and the trial jury are pure fan-out: spawn, collect,153merge. Codex uses subagents when the user has explicitly authorized parallel154agent work. Each reviewer or juror gets only the quoted manuscript context and a155strict JSON contract; no peer report, ledger, prior round, or project file path is156included in the prompt.157158But the loop has genuine human gates (the author reviews the issue list, gives159per-issue direction, authorizes edits, breaks ties). Fan-out phases run to160completion and return a result; they do not pause mid-run for hours of human161input. So:162163- fan-out steps (reviewers, trial, polish, recall, merge) -> **Codex subagents**164- human gates (per-issue direction, authorization, override) -> **main conversation turns**165- cross-round truth (the ledger) + stable conventions -> **project-owned memory files or host memory**166167## Review mode: one round, end to end168169The full adversarial loop (the v3 courtroom engine). Use it to harden the paper,170not for a single asked-for edit (that is direct-edit mode). Full protocol + the17114 orchestrator seams: `references/review-engine-v3.md`. `[SF]` = semantic172fan-out step using Codex subagents per `codex/runtime.md` and173`codex/phase-contracts.md`; `[det]` = deterministic Node guard run174orchestrator-side between semantic steps; `[HUMAN]` = author gate; `[LEDGER]` =175state write.1761771. **Resolve + recall.** Resolve the inputs above; recall this paper's conventions178 from memory. Pick scope: `full` (whole paper) or `passage` (one section / para / claim).1792. **`[det]` decompose.** Split the manuscript into reading units + stable180 `passage_id`s + the canonical section list.1813. **`[SF]` assign-reviewers** + **`[HUMAN]` confirm.** Name N subfields (2-4,182 default 3); instantiate N holistic domain reviewers from the gatekeeper core + a183 generated overlay. An unconfirmable slot degrades per slot to a generic gatekeeper184 (the three generic lenses in `reviewer-personas.md` are the fallback). The author185 confirms the assignment (or pins it via config).1864. **`[SF]` reading-check.** Each reviewer reads the WHOLE paper → weaknesses187 {`significance`(major|minor), `kind`(mechanical|substantive), verbatim quote —188 cannot quote = did not read} + one `overall_confidence` + a per-section coverage189 report. Anti-skim is three layers: `[det]` per-section quote-verify, `[SF]`190 coverage-auditor, `[SF]` targeted re-invoke.1915. **`[SF]` merge.** Semantic dedup across reviewers; derive `significance` (MAX) /192 `kind` (substantive-dominates) / corroboration. `[LEDGER]` intake as `raised`.1936. **`[det]` route.** mechanical → polish; substantive&minor → polish;194 substantive&major → trial (two parallel tracks).1957. **`[SF]` trial.** Per substantive-major charge: a whole-paper DEFENSE → 5196 decorrelated local-context jurors (+ on-demand expansion) → a deterministic verdict197 (decide iff quorum `surviving >= ceil(0.8*jurySize)` AND one side `> 60%` of198 surviving votes; else escalate to 12). Verdict ∈ {invalid-drop, valid-fixable,199 author-required, escalate}; the judge sets a `close_criterion` ONLY for a200 valid-fixable charge, satisfiable by editing existing text (no new data). `[SF]`201 polish runs the off-gate mechanical/minor track in parallel (never silently dropped).2028. **`[SF]` recall-audit.** Mode A revives wrongly-dropped charges; Mode B spot-checks203 strong-consensus majors BEFORE the edit. Runs before the drafter.2049. **`[HUMAN]` Authorize + `[SF]` drafter + edit-safety.** On authorization, the205 drafter writes the minimal patch per surviving valid-fixable. The edit-safety chain206 gates it: `[det]` anchor-diff + cross-ref → `[SF]` meaning-audit (frozen anchor,207 four-state) / edit-audit (risky non-anchor); `[det]` apply-patch + compile-guard land208 a passing patch and `[LEDGER]` mark `closed`; a drift / anchor / failed edit is209 reverted and queued. Revision logs / back-translations stay author-side.21010. **`[SF]` clerk + report.** The clerk reconciles the round boundary (carried211 open-questions vs this round's edits, via a passage_id + similarity merge key) and212 emits convergence counts. Summarize new/closed counts; in review mode do not213 auto-start the next round (auto mode drives the outer loop via `/goal`).214215GATE: `node scripts/ledger.js gate` = 0 gate-blocking active major (gate-blocking =216{raised, in-trial, re-trial, valid-fixable}; author-required / queued / dropped /217closed are gate-OK and author-required accumulates to the queue). Full protocol +218ledger schema + status machine: `references/review-engine-v3.md`,219`references/ledger-schema.md`. The single-pass 3-reviewer panel is available as220the `review-panel` fast path in `codex/phase-contracts.md`.221222## Hard rules (load-bearing, venue-agnostic)2232241. **Never edit the manuscript without explicit author sign-off.** Auto-mode225 carve-out: the rule HOLDS; auto satisfies it via UP-FRONT sign-off (the spine226 confirmation + the pre-authorized bounded-aggressive policy) plus the return227 queue, not per-edit sign-off. Nothing outside the authorized envelope is applied.2282. **Reviewers / jurors are isolated.** Fresh eyes per round: no cross-talk, no229 prior-round leakage, no sight of the ledger. Enforced by (a) what goes into each230 agent's prompt AND (b) an explicit ISOLATION instruction in every reviewer-type231 prompt telling the agent to judge only the quoted text and not read files.2323. **A valid-fixable issue carries a `close_criterion`** (one concrete sentence an233 edit must satisfy), set by the judge at trial; it is null at intake.2344. **No leakage into the reviewed text.** Revision logs, back-translations, and235 self-check verdicts are author-side aids; they never enter the manuscript or236 any frozen snapshot.2375. **Disagreement resolves through discussion, then override** (logged), never a238 silent dismissal.2396. **No hardcoded paths or project files in the skill.** Resolve at runtime.240241## Memory convention242243- At round start: recall the paper's conventions (house style, venue, persona244 tuning) from memory; read the resolved `LEDGER.json` for open issues.245- During the round: the ledger is the only mutable truth; update it at merge,246 trial verdicts, recall, and close.247- After the round: persist any newly learned stable convention to memory (e.g. a248 house-style rule a reviewer surfaced), not the transient issue state.249250## Intensity and host-specific launch notes251252The quick panel can run in stronger or cheaper forms using the `review-panel`253fast path in `codex/phase-contracts.md`:254255- **loop-until-dry**: re-runs independent fresh panels and accumulates only issues256 not seen before, stopping after `dryStop` consecutive passes that add no257 surviving issue (hard cap `maxRounds`). Raises recall past a single pass.258- **adversarial verify**: each new issue faces perspective-diverse skeptics259 (misreading / already-addressed / scope-or-severity) and is kept unless a260 majority refute it, filtering plausible-but-wrong issues before they reach the261 ledger. Bias is to keep, so real flaws are not lost.262263Toggle via args: thorough mode -> defaults (`maxRounds` 4, `dryStop` 2,264`verify` true); light mode -> pass `{maxRounds:1, verify:false}` for the basic265single-panel form. The loop is budget-aware and stops early if the token budget266runs low. Codex treats intensity as an ordinary runtime setting, not as a267separate launch keyword.268269270271## Capabilities and status272273Built: the review engine; the submission-readiness checker (deterministic desk-reject screening plus a real LaTeX compile, degrading to a structural lint when no toolchain is present); and auto mode (the review-revise loop toward a goal under a drift-bounded policy, applying safe fixes and queueing risky ones for author review). Roadmap: vision-based layout verification, automatic venue detection from the class file, and reviewer personas tuned to each venue community.274275## Related skills276277- `ml-paper-writing`: from-scratch drafting, citation verification (never278 hallucinate citations), conference checklists. This loop borrows its279 sentence-level guidance for the edit-drafting step rather than duplicating it.280- `academic-plotting`: figure and architecture-diagram generation (out of scope281 here; this loop edits text and captions, not figure images).