paperlint 6.3.0 → 7.1.0

package/presets/aisec.md

@@ -0,0 +1,117 @@
+---
+title: "AISec @ ACM CCS — venue card"
+---
+
+# AISec @ ACM CCS
+
+The card of the venue preset `paperlint:aisec` ([`aisec.jsonc`](aisec.jsonc)): what the venue asks
+for, in prose. The numbers a machine checks live in the preset; the section at the end is generated
+from it. The `submit-paper` skill reads this card for the venue's specifics.
+
+**Re-verify the call for papers every year** ([aisec.cc](https://aisec.cc/)) — deadline, page limit
+and tracks shift between editions.
+
+## The bar
+
+AISec is the long-running AI-security workshop of **ACM CCS**, a top security conference — not a
+first-edition workshop. Mature threat modelling, honest evaluation and sound ethics are expected,
+not rewarded: a security committee rejects on soundness and ethics grounds that a software
+engineering workshop would wave through.
+
+## The venue (2026 edition)
+
+- **Venue**: ACM Workshop on Artificial Intelligence and Security (**AISec**), co-located with
+  **ACM CCS** (Computer and Communications Security).
+- **Deadline**: 24 July 2026 (check the AoE cutoff on the call each cycle).
+- **Tracks**: full research papers; **benchmark papers** (new evaluation frameworks or datasets
+  **with artifact sharing** — a track of its own); position and SoK (systematization of knowledge)
+  papers. Pitch to the bar of the chosen track.
+- **Format**: ACM double-column (`sigconf`). Build with `pdflatex; bibtex; pdflatex; pdflatex`; add
+  `\microtypesetup{expansion=false}` (the most common acmart crash).
+- **Blind model**: **double-blind** — `\documentclass[sigconf,review,anonymous]{acmart}`, author
+  `Anonymous Author(s)`. See `submit-paper` §0/§4.
+- **Archival**: accepted papers go into the CCS workshop proceedings in the ACM Digital Library,
+  with a DOI.
+- **No supplementary-material field** on the form: host the artifact externally, anonymized (a
+  view-only link), and `\url{}` it in the Availability section. See `submit-paper` §2 and
+  `paper-pipeline/references/anonymization.md`.
+
+## Paper kinds and limits
+
+Verbatim from the call: «Submitted papers must be at most 10 pages long in double-column ACM
+format---excluding bibliography and well-marked appendices, for which up to two additional pages
+can be used (hence, the paper's overall length must be of 12 pages at most).»
+
+The limits are the same for every track; the preset keeps one kind per track so a paper's
+`paperlint.json` says which track it was submitted under.
+
+| kind        | body pages | pages of bibliography and appendices |
+| ----------- | ---------: | -----------------------------------: |
+| `research`  |         10 |                                    2 |
+| `benchmark` |         10 |                                    2 |
+| `position`  |         10 |                                    2 |
+| `sok`       |         10 |                                    2 |
+
+⚠️ The venue tolerates a 13th page that holds **only** the generative-AI paragraph («Too many pages:
+the limit is 12 pages, found 13» is then acceptable). `format/page-limit` does not know this
+exception and reports the 13th page; a person clears that case by eye.
+
+## What else the preset encodes
+
+- **The ACM template**, through `extends: paperlint:acm-sigconf` — AISec points to the main ACM
+  CCS formatting instructions and publishes no numbers of its own ([`acm-sigconf.md`](acm-sigconf.md)).
+- **Names**: `AISec` and `ACM CCS`, what `tex/venue-leftover` looks for in another venue's paper.
+- **No rules of its own**: nothing says who produces the AISec proceedings or whether they require
+  a balanced last page. A paper turns `pdf/last-page-balance` on in its own `paperlint.json` if the
+  camera-ready instructions ask for it.
+
+## What decides acceptance at a security venue
+
+1. **Threat-model soundness.** Name the attacker and their capabilities (what they observe and
+   control, their goal, their budget), and separate accidental failure from an adaptive adversary
+   — conflating the two is the classic reject. The threat model goes up front.
+2. **Ethics and responsible disclosure — a hard reject vector.**
+   - Anonymize studied third parties too: aggregate statistics and use anonymized IDs for the
+     entities studied, rather than naming maintainers or projects.
+   - Disclose findings to affected parties **before** any de-anonymized release; the paper states
+     the disclosure timeline and the release gate that holds de-anonymized data back until then.
+   - The anonymized artifact must itself be leak-free — the released files, not just the PDF. A
+     named entity in the artifact is invisible to a paper-only read.
+3. **Dual use.** A paper that publishes an evasion or attack technique publishes its fix with it.
+4. **Generative-AI disclosure.** ACM and CCS require stating how generative AI was used in the
+   work (writing, code, experiments).
+
+## Track fit
+
+Adversarial ML, ML for security and security for ML, LLM and agent safety and security, evaluation
+and benchmarking of AI-security defenses, poisoning, evasion and extraction, and rigorous
+measurement of whether a claimed safety mechanism holds. A measurement paper that ships a reusable
+evaluation framework and an anonymized artifact fits the benchmark track; name the track in the
+abstract.
+
+## Compose with
+
+- `submit-paper` — the mechanics: HotCRP form, anonymous artifact hosting, "ready for review".
+- `camera-ready` — de-anonymization and the release gate after acceptance.
+- `pc-panel-review` — its ethics lens and artifact-runner lens (plant a leak, confirm it is caught).
+- `paper-pipeline/references/artifact-checklist.md` and `paper-pipeline/references/anonymization.md`.
+
+## Camera-ready
+
+AISec is an ACM venue, so the final's mechanics are the ACM ones in the `submit-paper` skill's
+`references/publishers/acm.md`: eRights first (it needs the final title), the preamble out of
+submission mode, source files beside the PDF, CCS concepts from the 2012 CCS browser. AISec's own
+DOI, ISBN, copyright line and dates are added here when the venue publishes them.
+
+## What the preset resolves to
+
+<!-- paperlint:preset-rules -->
+<!-- prettier-ignore-start -->
+<!-- Generated from aisec.jsonc by `npx eslint --fix` (rule preset/card-rules). Edit the preset, not this section. -->
+
+**`paperlint:aisec`** — venue preset for AISec · call for papers: <https://aisec.cc/> · extends `paperlint:acm-sigconf`
+
+No preset of this chain sets a rule: under it, every rule keeps paperlint's default.
+
+<!-- prettier-ignore-end -->
+<!-- /paperlint:preset-rules -->

package/presets/ieee-conference.jsonc

@@ -24,7 +24,12 @@
 // aidc.jsonc). The planted defects fail as they should, measured on the same compsoc paper:
 //   `a4paper` added to the options        → page 8.264 x 11.694 in: pdf/geometry
 //   Computer Modern (`\rmdefault{cmr}`)   → fonts CMR10, CMBX12, …: pdf/fonts, no NimbusRomNo9L
+//
+// Its card, in prose: `ieee-conference.md` beside this file.
 {
+  // A template family: it names no venue, so no `name` and no `url` (the venues that extend it do).
+  "type": "family",
+
   "template": "\\documentclass[conference]{IEEEtran}",
 
   // TeX Live packages the IEEEtran template needs ON TOP OF tex-base.jsonc, each with the files that

package/presets/ieee-conference.md

@@ -0,0 +1,38 @@
+---
+title: "IEEE IEEEtran conference — template family card"
+---
+
+# IEEE `IEEEtran` conference
+
+The card of the template family `paperlint:ieee-conference`
+([`ieee-conference.jsonc`](ieee-conference.jsonc)). A family is what a publisher's template
+decides, whatever the venue: it names no venue and has no call for papers. A venue preset extends
+it and adds what its own call sets — the class options it names, the page limit of each kind of
+paper, the sections it requires. The template is [IEEEtran on CTAN](https://ctan.org/pkg/ieeetran).
+
+## What the family encodes
+
+- **The class**: `\documentclass[conference]{IEEEtran}`. A venue that requires `compsoc` names it
+  in its own template and its own text block (as `aidc` does).
+- **The page**: US letter, 8.5 × 11 in; the text block of `[conference]`, 7.14 × 9.30 in; two
+  columns.
+- **The fonts**: Times (the URW clone the PDF names NimbusRomNo9L) for text, headings and title;
+  body at 10 pt, the bibliography at 8 pt (IEEEtran sets it in `\footnotesize`).
+- **TeX packages** on top of the base set: ieeetran, cite, psnfss and times.
+
+The numbers are measured, not copied: IEEE's own template files (`bare_conf.tex`,
+`bare_conf_compsoc.tex`) and a 13-page paper built on each with IEEEtran.cls V1.8b, on two TeX Live
+versions. `compsoc` changes the text block and the headings, and nothing the `pdf/*` rules judge.
+
+## What the family resolves to
+
+<!-- paperlint:preset-rules -->
+<!-- prettier-ignore-start -->
+<!-- Generated from ieee-conference.jsonc by `npx eslint --fix` (rule preset/card-rules). Edit the preset, not this section. -->
+
+**`paperlint:ieee-conference`** — template family, no venue of its own · extended by `paperlint:aidc`, `paperlint:msr`
+
+No preset of this chain sets a rule: under it, every rule keeps paperlint's default.
+
+<!-- prettier-ignore-end -->
+<!-- /paperlint:preset-rules -->

package/presets/msr.jsonc

@@ -0,0 +1,71 @@
+// MSR 2027 Technical Papers (the 24th International Conference on Mining Software Repositories) — a
+// VENUE preset over the IEEE conference family. Everything the IEEEtran template decides (page
+// size, columns, fonts, body size, TeX packages) comes from ieee-conference.jsonc; this file adds
+// what MSR's call for papers sets.
+//
+// Verbatim from https://2027.msrconf.org/track/msr-2027-technical-papers (fetched 2026-10-05):
+//   «All submissions must conform to the IEEE conference proceedings template, specified in the
+//    IEEE Conference Proceedings Formatting Guidelines (title in 24pt font and full text in 10pt
+//    type, LaTeX users must use \documentclass[10pt,conference]{IEEEtran} without including the
+//    compsoc or compsocconf options).»
+//   «Note that IEEE format is being used this year, whereas last year it was ACM format.»
+//   «All submissions must not exceed 10 pages for the main text, inclusive of all figures, tables,
+//    appendices, etc. Two more pages containing only references are permitted. All submissions
+//    must be in PDF. Accepted papers will be allowed one extra page for the main text of the
+//    camera-ready version.»
+//   «The MSR 2027 Technical Track will employ a double-anonymous review process.»
+//
+// The dates are not lint facts; they are in the card, msr.md beside this file.
+{
+  "type": "venue",
+  "name": "MSR",
+  // The call quoted above.
+  "url": "https://2027.msrconf.org/track/msr-2027-technical-papers",
+  "extends": "paperlint:ieee-conference",
+
+  // «The MSR 2027 Technical Track will employ a double-anonymous review process. Thus, no
+  // submission may reveal its authors' identities.» (the call above), and `"blind":true` in the
+  // portal's data (https://msr2027.hotcrp.com/deadlines, 2026-10-05). anonymity/identity then
+  // requires the paper's `identity`.
+  "blind": true,
+
+  // «Papers must be submitted through HotCRP: https://msr2027.hotcrp.com/» (the call above).
+  // Recorded 2026-10-05: `/api/whoami` on this site answers 401 "Missing credentials" without a
+  // token — the HotCRP API, there.
+  "portal": { "kind": "hotcrp", "url": "https://msr2027.hotcrp.com" },
+
+  // What the venue is called in a paper's text (tex/venue-leftover) — and only beside a year
+  // («MSR 2027», «MSR'27»): «MSR» is also Microsoft Research, and an acronym papers define for
+  // themselves («Manual Speech Recognition (MSR)», in an accepted ACSAC paper of
+  // fixtures/accepted-papers/, warned six times when any whole word counted).
+  "aliases": ["MSR"],
+  "mentions": "with-year",
+
+  // The class line the call names, with the two options it forbids (tex/template). The family's
+  // `[conference]` text block (7.14 x 9.30 in) is the one this line builds, so no `format` override.
+  "template": "\\documentclass[10pt,conference]{IEEEtran}",
+  "template_forbids": ["compsoc", "compsocconf"],
+
+  "format": {
+    // 🔴 THE MAIN TEXT INCLUDES THE APPENDICES, AND THE REFERENCE PAGES ARE THE ONES WITH ONLY
+    // REFERENCES — not banal's count. Measured 2026-10-05 on IEEEtran [10pt,conference] builds:
+    //   10 pages of main text, its appendix before the references, 2 reference-only pages
+    //     → banal: body 9, references 3 (a page whose first heading reads «Appendix» is typed as
+    //       bibliography) — a correct paper fails
+    //   11 pages of main text, the appendix after the references
+    //     → banal: body 10, references 2 (the appendix taken for more references) — an over-long
+    //       paper passes
+    // So format/page-limit counts from the text (`appendix_in_body`): the body is every page up to
+    // the references, that page too when body text stands above the heading, plus every page from
+    // an appendix that follows the references; the pages between are the reference pages
+    // (docs/rules/format/page-limit.md). An appendix after the references is found by hyperref's
+    // anchor: load hyperref.
+    "body_ends_at": "references",
+    "appendix_in_body": true,
+
+    // One kind: the Technical Track's papers. The camera-ready's extra page is not this limit.
+    "kinds": {
+      "technical": { "body_pages_max": 10, "ref_pages_max": 2 },
+    },
+  },
+}

package/presets/msr.md

@@ -0,0 +1,94 @@
+---
+title: "MSR — venue card"
+---
+
+# MSR (Technical Papers)
+
+The card of the venue preset `paperlint:msr` ([`msr.jsonc`](msr.jsonc)): what the venue asks for,
+in prose. The numbers a machine checks live in the preset; the section at the end is generated from
+it. The `submit-paper` skill reads this card for the venue's specifics.
+
+**Re-verify the call for papers every year**
+([2027.msrconf.org/track/msr-2027-technical-papers](https://2027.msrconf.org/track/msr-2027-technical-papers)):
+the template changed from ACM to IEEE in 2027 («Note that IEEE format is being used this year,
+whereas last year it was ACM format.»), and limits and dates change between editions.
+
+## The venue (2027 edition)
+
+- **Venue**: the Technical Papers track of **MSR 2027**, the International Conference on Mining
+  Software Repositories.
+- **Portal**: HotCRP, `https://msr2027.hotcrp.com/` («Papers must be submitted through HotCRP»).
+- **Dates** (the call, «AoE (UTC-12h)»): abstract **2026-10-20**, paper **2026-10-23**; early
+  reject notification 2026-12-03; author response 2026-12-04 to 2026-12-08; notification
+  2027-01-08; camera-ready 2027-01-26. The call's prose names the weekdays «Mon, 20 Oct 2026» and
+  «Thurs, 23 Oct 2026», its date table «Tue 20 Oct 2026» and «Fri 23 Oct 2026»; the calendar sides
+  with the table (2026-10-20 is a Tuesday).
+- 🔴 **The portal closes earlier than the call says.** Read 2026-10-05 on
+  `https://msr2027.hotcrp.com/deadlines`: «Submission deadline: Tuesday Oct 20, 2026, 12 AM EDT»
+  (`"sub":1792468800`, 2026-10-20 04:00 UTC) and «Resubmission deadline: Friday Oct 23, 2026, 12 AM
+  EDT» («Completed submissions may be updated until this deadline»). Midnight EDT at the start of
+  each day is about a day and a half before the end of that day AoE. The portal is what refuses the
+  upload: plan for its times, and re-read the page, since chairs change them.
+- **Format**, verbatim from the call: «LaTeX users must use \documentclass[10pt,conference]{IEEEtran}
+  without including the compsoc or compsocconf options», «title in 24pt font and full text in 10pt
+  type», and «Alterations of spacing, font size, and other changes that deviate from the
+  instructions may result in desk rejection without further review.»
+- **Double-anonymous**: «The MSR 2027 Technical Track will employ a double-anonymous review process.
+  Thus, no submission may reveal its authors' identities.» The call adds: names omitted, prior work
+  cited in the third person, and a preprint must not say it was submitted to MSR. The portal's data
+  says `"blind":true`.
+- **Attendance**: «If a submission is accepted, at least one author of the paper is required to
+  register for MSR 2027 and present the paper.» The call does not say whether a remote presentation
+  is allowed.
+- **New in 2027**, from the call: a submission may be desk rejected for fabricated references
+  («citations to papers, authors, venues, datasets, tools, or other artifacts that do not exist»),
+  clear signs of unverified AI-generated text, or hidden instructions aimed at automated tools.
+
+## Paper kinds and limits
+
+| kind        | main-text pages | pages of only references | what the call says                                                                                                                                                        |
+| ----------- | --------------: | -----------------------: | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `technical` |              10 |                        2 | «All submissions must not exceed 10 pages for the main text, inclusive of all figures, tables, appendices, etc. Two more pages containing only references are permitted.» |
+
+«Accepted papers will be allowed one extra page for the main text of the camera-ready version.» —
+not the submission's limit, so not in the preset.
+
+🔴 **Appendices count in the ten pages**, and a page that holds anything but references is a
+main-text page. banal's count gets both wrong on an IEEEtran build (an appendix before the
+references turns its pages into «bibliography»; one after them into «references»), so the preset
+counts from the text (`format.appendix_in_body`): the main text is every page up to the references —
+that page too when text stands above the heading — and every page from an appendix that follows the
+references; the pages between hold only references. **Load hyperref**: an appendix after the
+references is found by hyperref's anchor, and without it is counted as reference pages. See
+`docs/rules/format/page-limit.md`.
+
+## What else the preset encodes
+
+- **The IEEE template**, through `extends: paperlint:ieee-conference` — letter paper, two columns,
+  Times at 10 pt, IEEEtran's TeX packages ([`ieee-conference.md`](ieee-conference.md)) — with
+  `10pt` and `conference` required in the class line and `compsoc` and `compsocconf` forbidden
+  (`tex/template`, `template_forbids`).
+- **Double-blind review** (`blind`), so `anonymity/identity` checks the PDF against the paper's
+  declared `identity`.
+- **Where submissions go** (`portal`): `paperlint submission` reads and updates a paper's
+  submission through HotCRP's API on that site.
+- **Names**: `MSR`, what `tex/venue-leftover` looks for in another venue's paper — only with a year
+  beside it (`mentions: with-year`: «MSR 2027», «MSR'27»), since «MSR» is also Microsoft Research
+  and an acronym papers define for themselves.
+- **Not encoded**: the capitalization of the title and headings (`tex/heading-case` stays off: the
+  IEEE guidelines' rule has not been read and quoted here), and the register of the body (no
+  accepted MSR paper in the IEEE template exists yet to anchor `tex/contrast-frames` and its
+  siblings).
+
+## What the preset resolves to
+
+<!-- paperlint:preset-rules -->
+<!-- prettier-ignore-start -->
+<!-- Generated from msr.jsonc by `npx eslint --fix` (rule preset/card-rules). Edit the preset, not this section. -->
+
+**`paperlint:msr`** — venue preset for MSR · call for papers: <https://2027.msrconf.org/track/msr-2027-technical-papers> · extends `paperlint:ieee-conference`
+
+No preset of this chain sets a rule: under it, every rule keeps paperlint's default.
+
+<!-- prettier-ignore-end -->
+<!-- /paperlint:preset-rules -->

package/presets/realm.jsonc

@@ -1,5 +1,5 @@
-// REALM @ EMNLP 2026 format profile — ACL family, NOT ACM. Prose is in `realm.md` nearby,
-// the family's shared rules are in `paper-pipeline/references/acl-venue-rules.md`.
+// REALM @ EMNLP 2026 format profile — ACL family, NOT ACM. Prose is in its card, `realm.md`
+// beside this file; the family's shared rules are in `paper-pipeline/references/acl-venue-rules.md`.
 //
 // 🔴 THE GEOMETRY NUMBERS ARE MEASURED, NOT COPIED. banal run against our own build
 // (`repro/build-submission.sh` → `build/acl_latex.pdf`, 2026-08-26):
@@ -8,6 +8,11 @@
 // acmart (letter 8.5x11, size 9, Libertine) shares NOTHING with this — copying the numbers from
 // `agenticdev.jsonc` would give a gate that reddens on a correct paper.
 {
+  "type": "venue",
+  "name": "REALM",
+  // The workshop's call for papers, which the page limits below are quoted from.
+  "url": "https://realm-workshop.github.io/call_for_papers",
+
   // ACL's template is not a class: `acl_latex.tex` opens with `\documentclass[11pt]{article}` and
   // loads `acl.sty` as a package (https://github.com/acl-org/acl-style-files, checked 2026-09-27).
   // So the class `tex/template` compares against is article, at 11pt.

package/presets/realm.md

@@ -0,0 +1,166 @@
+---
+title: "REALM @ EMNLP — venue card"
+---
+
+# REALM @ EMNLP
+
+The card of the venue preset `paperlint:realm` ([`realm.jsonc`](realm.jsonc)): what the venue asks
+for, in prose. The numbers a machine checks live in the preset; the section at the end is generated
+from it. The `submit-paper` skill reads this card for the venue's specifics.
+
+**Family: ACL.** What is ACL-wide (the appendix rule, Limitations and Ethics outside the limit,
+anonymity, OpenReview mechanics) lives in `paper-pipeline/references/acl-venue-rules.md`; only
+REALM's own facts are here. ACL's template is not a class — `acl_latex.tex` loads `acl.sty` into
+`\documentclass[11pt]{article}` — so this preset extends no family and carries the template's
+numbers itself.
+
+Sources: the call for papers,
+[realm-workshop.github.io/call_for_papers](https://realm-workshop.github.io/call_for_papers)
+(fetched 2026-08-02 and 2026-08-03), and the OpenReview invitation schema.
+🔴 **Re-verify every cycle** — dates, limits and tracks move between editions.
+
+## The venue (2026 edition, the 2nd)
+
+| field                      | value                                                                                                         |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------- |
+| full name                  | REALM — workshop at EMNLP 2026                                                                                |
+| edition                    | **2nd** (the 1st edition's accepted papers exist — mine them with `study-accepted-papers`)                    |
+| direct submission deadline | **Wednesday 5 August 2026**, 23:59 **AoE** (= 6 August 11:59 UTC)                                             |
+| ARR commitment deadline    | 31 August 2026, same AoE convention                                                                           |
+| archival                   | optional at submission; archival → **ACL Anthology proceedings**                                              |
+| blind                      | **double-blind** — names and affiliations omitted                                                             |
+| preprints                  | allowed at any time, provided the PDF does not link to de-anonymized information                              |
+| portal                     | `openreview.net/group?id=EMNLP/2026/Workshop/REALM`                                                           |
+| format                     | **PDF only**                                                                                                  |
+| mandatory sections         | none beyond ACL's Limitations; _"authors are expected to attest to the ethical considerations in their work"_ |
+| venue mode                 | hybrid; Thursday 29 October, 9:00–17:30, Room P1                                                              |
+
+## Paper kinds and limits
+
+| kind    | at submission | at camera-ready | references and appendices |
+| ------- | ------------: | --------------: | ------------------------: |
+| `long`  |       8 pages |         9 pages |                 unlimited |
+| `short` |       4 pages |         5 pages |                 unlimited |
+
+🔴 **The page limit is not gated by the preset**, and that is a measurement: banal counts
+Limitations and Ethics as body, which ACL excludes from the limit, so a gate on its count reports a
+paper inside the limit. The numbers sit in the preset's `page_limits_not_gated`, as data for a
+counter that knows where the body ends; the two kinds exist so a paper can name its own.
+
+## What else the preset encodes
+
+- **The template**: `\documentclass[11pt]{article}` (`tex/template`), A4, two columns, Times at
+  11 pt, references between 9 and 11 pt — measured on a build, not copied from ACL's text.
+- **TeX packages** `acl.sty` needs: lineno, and the Times, Helvetica, Courier, Symbol and Zapf
+  Dingbats fonts (without them: "Font ptmr8t not loadable" and no PDF).
+- **Names**: `REALM` and `EMNLP`, what `tex/venue-leftover` looks for in another venue's paper.
+
+## 🔴 The submission form has no supplementary field
+
+From `api2.openreview.net/invitations?id=EMNLP/2026/Workshop/REALM/-/Submission`, the form's
+**complete** field list:
+
+```
+title · authors · authorids · keywords · TLDR · abstract · pdf ·
+archival · cross_submission_to · serve_as_reviewer · venue · venueid
+```
+
+There is no Responsible NLP Checklist and no supplementary upload. So:
+
+1. **Host the artifact externally and anonymously**, with the URL inside the PDF — a submission
+   blocker, and an artifact reviewer refuses the Available badge without a resolving URL.
+2. **Do not confuse this route with ARR**, whose portal does have the checklist and separate
+   software and data uploads.
+3. **No checklist means no LLM-disclosure field here**: put the disclosure in Limitations or Ethical
+   considerations, both outside the page limit
+   (`paper-pipeline/references/acl-venue-rules.md`, §GenAI / LLM disclosure).
+
+`keywords` is a form field, so a keyword line does **not** belong in the typeset PDF.
+
+## The submission form, field by field
+
+What the live form shows, in order. Help text is quoted.
+
+| #   | field                   | required | type                                         | what goes in                                                                                                                                                                                                                                   |
+| --- | ----------------------- | -------- | -------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| 1   | **Title**               | ✱        | one line                                     | _"Add TeX formulas using $In-line$ or $$Block$$"_ — plain text otherwise. **Must match the PDF's title exactly**; a late title edit in the paper means re-uploading the PDF _and_ retyping this                                                |
+| 2   | **Authors**             | ✱        | profile search                               | pre-filled with the submitting author. _"All authors must have an OpenReview profile prior to submitting"_ — a co-author without an active profile blocks the whole submission                                                                 |
+| 3   | **Keywords**            | ✱        | comma-separated                              | free text, no controlled vocabulary. 🔴 **Because this is a form field, the keyword line does NOT belong in the typeset PDF**                                                                                                                  |
+| 4   | **TL;DR**               | —        | one sentence, 🔴 **hard max 250 characters** | _"a short sentence describing your paper"_. The limit is **not in the help text** — it rejects on submit. Optional, but it is the first thing a bidding reviewer reads: name the mechanism, not the topic. Count the characters before pasting |
+| 5   | **Abstract**            | ✱        | markdown + TeX, Write/Preview tabs           | paste the paper's abstract. Markdown emphasis renders; `**bold**` survives. Backticks render as code                                                                                                                                           |
+| 6   | **PDF**                 | ✱        | file upload                                  | _"Upload a PDF file that ends with .pdf"_. No supplementary field anywhere on the form — see above                                                                                                                                             |
+| 7   | **Archival**            | ✱        | radio: Archival / Non-archival               | Archival → **ACL Anthology**. _"If your paper is a cross-submission (currently under review or already published at another venue), it cannot be archival, per ACL policies."_ Changeable at camera-ready                                      |
+| 8   | **Cross Submission To** | —        | one line                                     | leave **empty** unless the paper is genuinely under review elsewhere. Filling it forfeits archival                                                                                                                                             |
+| 9   | **Serve As Reviewer**   | ✱        | profile search                               | _"reciprocal reviewing practice… nominate at least one author to serve as a reviewer using their OpenReview profile ID (e.g. ~First_Last1)"_. **A solo author nominates themself** — the field is required and there is nobody else            |
+| 10  | License                 | ✱        | fixed chip                                   | **CC BY 4.0**, not editable                                                                                                                                                                                                                    |
+| 11  | Readers                 | ✱        | fixed chips                                  | `EMNLP 2026 Workshop REALM` + `authorids`                                                                                                                                                                                                      |
+| 12  | Signatures              | ✱        | fixed chip                                   | the submitting author                                                                                                                                                                                                                          |
+
+Then **Submit** / Cancel.
+
+🔴 **The displayed deadline is not when writes stop.** An OpenReview invitation carries two dates
+and the interface shows only the first:
+
+| field                                                          | 2026 edition, UTC |
+| -------------------------------------------------------------- | ----------------- |
+| `duedate` — printed everywhere as the deadline                 | 2026-08-06 11:59  |
+| `expdate` — when the invitation actually stops accepting edits | 2026-08-06 12:29  |
+
+A 30-minute window, undocumented in the call and invisible on the form. Read it before the last
+hour:
+
+```
+curl -sS "https://api2.openreview.net/invitations?id=<VENUE-PATH>/-/Submission" \
+  | python3 -c "import json,sys,datetime;[print(k, datetime.datetime.utcfromtimestamp(i[k]/1000)) \
+    for i in json.load(sys.stdin)['invitations'] for k in ('duedate','expdate') if i.get(k)]"
+```
+
+Do not plan around it — a venue may set `expdate == duedate`. It is room to recover from a defect
+found at the deadline, not a schedule.
+
+## Traps
+
+- **The title lives in two places, and they diverge silently.** Change the paper first, rebuild,
+  then fill the form — never the other way round.
+- **"Serve As Reviewer" enrols the nominee.** The day after submission the program chairs assign
+  papers and set a review deadline (2026: submissions 5 August, reviews due 21 August AoE, sixteen
+  days). ACL's AI policy applies to the reviews, verbatim from the chairs' email: _"the use of
+  generative AI tools to write an entire review is prohibited. While such tools may be used for
+  minor editing (e.g., grammar or clarity), the content and assessment must be the reviewer's
+  own."_ The sanction lands on submissions to the next ARR cycle. Price the reviewing window in
+  before submitting, and check whether a decision about reviewing was already made before filling
+  a required field.
+- **Keywords are a form field, not paper content.** Keep the list in a non-typeset comment in the
+  source and paste it here.
+- **The abstract field accepts markdown**, so paste the source abstract. Extracting it from the
+  built PDF is a mistake: on a line-numbered build the extractor interleaves the margin numbers and
+  drops paragraphs.
+- **The deadline's time zone is stated in a different sentence** of the call than the date. Read
+  the whole page, then convert.
+- **OpenReview profile moderation is the critical path.** An author with no institutional email and
+  no DBLP record has no expedite route; start the profile weeks ahead.
+- The workshop's hybrid status comes from REALM's own root page, not EMNLP's general one.
+
+## The ARR route is not a free fallback
+
+The 31 August "ARR commitment" deadline commits a paper **already reviewed by ACL Rolling Review**.
+A paper never submitted to ARR needs an ARR submission first, on ARR's own cycle.
+
+## Sources
+
+- `realm-workshop.github.io/call_for_papers` — the call (fetched 2026-08-02, 2026-08-03)
+- `realm-workshop.github.io/` — the root page, hybrid confirmation
+- `api2.openreview.net/invitations?id=…/-/Submission` — the live form schema
+
+## What the preset resolves to
+
+<!-- paperlint:preset-rules -->
+<!-- prettier-ignore-start -->
+<!-- Generated from realm.jsonc by `npx eslint --fix` (rule preset/card-rules). Edit the preset, not this section. -->
+
+**`paperlint:realm`** — venue preset for REALM · call for papers: <https://realm-workshop.github.io/call_for_papers>
+
+No preset of this chain sets a rule: under it, every rule keeps paperlint's default.
+
+<!-- prettier-ignore-end -->
+<!-- /paperlint:preset-rules -->

package/presets/tex/agenticdev.tex

@@ -1,5 +1,5 @@
 % agenticdev.tex — venue numbers for the guards IN THE PREAMBLE. LaTeX doesn't read YAML, so the
-% values are duplicated from venues/agenticdev.jsonc (the source of truth — read by the
+% values are duplicated from presets/agenticdev.jsonc (the source of truth — read by the
 % `pdf/profile` rule, see eslint-rules/pdf-facts.mjs:128). in = inches.
 %
 % 🔴 THIS USED TO SAY `venues/agenticdev.yaml`, AND NO SUCH FILE EXISTS (fixed 2026-09-12). The

package/presets/tex-base.jsonc

@@ -14,7 +14,13 @@
 //   cm-super — without it pdflatex silently RASTERIZES Computer Modern in some contexts (a Type 3
 //     bitmap font in the PDF; measured 2026-09-01 on an ACL paper: 1 Type 3 font before, 0 after).
 //     ACM and ACL both reject such a PDF, and the build does not complain.
+//
+// Its card, in prose: `tex-base.md` beside this file.
 {
+  // The base set: not a venue and not a template family. No preset extends it — every paper gets it
+  // anyway, under whatever it extends — so it holds `tex` and nothing else (the schema says so).
+  "type": "base",
+
   "tex": {
     "packages": {
       "latex": ["latex.ltx", "article.cls"],

package/presets/tex-base.md

@@ -0,0 +1,30 @@
+---
+title: "The TeX base set — card"
+---
+
+# The TeX base set
+
+The card of [`tex-base.jsonc`](tex-base.jsonc): the TeX Live packages **every** paper gets, and all
+a paper gets when it extends no preset. It is neither a venue nor a template family, and no preset
+extends it — a venue's or a family's `tex.packages` is added on top of it, never instead of it.
+`paperlint toolchain` installs it with every preset's packages; `paperlint build` checks it is
+present before building.
+
+What it is for: a plain `article` paper with the usual tools, typeset in Type 1 fonts. Each package
+names the files that prove it is installed — `kpsewhich` must find them after an install, because
+an installer's exit code is not the criterion. `cm-super` is there because without it pdflatex
+silently rasterizes Computer Modern in some contexts, a bitmap font that ACM and ACL both reject.
+`texcount` and `checkcites` are programs the skills run, not files LaTeX loads.
+
+## What the preset resolves to
+
+<!-- paperlint:preset-rules -->
+<!-- prettier-ignore-start -->
+<!-- Generated from tex-base.jsonc by `npx eslint --fix` (rule preset/card-rules). Edit the preset, not this section. -->
+
+**`tex-base.jsonc`** — the TeX base set: every paper gets it, whatever it extends, and no preset extends it
+
+No preset of this chain sets a rule: under it, every rule keeps paperlint's default.
+
+<!-- prettier-ignore-end -->
+<!-- /paperlint:preset-rules -->