paperlint 2.0.0

package/skills/paper-pipeline/references/pipeline-status-template.md

@@ -0,0 +1,162 @@
+# PIPELINE-STATUS template — the per-paper readiness scorecard
+
+`paperlint new <name>` writes this into `<papers>/<name>/PIPELINE-STATUS.md` for each paper. It is the
+**durable, colocated, markdown source-of-truth** for _which pipeline work has run, its result, and what is still open_ — so
+"did everything run?" is a checked fact, not a memory. Chat evaporates; this file rides with the paper
+to camera-ready.
+
+🔴 **It is also machine-read.** `../scripts/pipeline-check.mjs` parses this file, so the four section
+headings, the row shape, and the ISO dates are a **format contract**, not decoration. Break them and
+the checks go quiet — which, as of 2026-08-03, is a failure mode this repo has lived through three
+times in one day. Run `node .claude/skills/paper-pipeline/scripts/pipeline-check.mjs <paper-dir>`
+after editing to confirm it still parses.
+
+## The front matter
+
+Above the sections, the file carries YAML. Two fields are read by rules:
+
+```yaml
+---
+stages:
+  - stage: submitted
+    date: 2026-07-22
+    pdf: versions/2026-07-22-submitted.pdf
+    bytes: 305412
+    source: versions/2026-07-22-submitted.tex
+    sourceBytes: 57210
+researchQuestion: "Does pruning the state space reduce review cost?"
+---
+```
+
+`researchQuestion` is **your sentence, written once** — not a label the checker hunts for in the
+prose. `paper/research-question` then asks two things it can actually answer: is it written down,
+and does the paper contain it (whitespace collapsed). Leave the field out and a shipped paper gets
+an advisory finding, which is the right outcome for a position paper with no question: the absence
+becomes a decision on the record rather than an omission.
+
+⚠️ Write the sentence the way it appears in the paper. If the paper breaks it across lines that is
+fine — only whitespace is normalised — but LaTeX markup inside the sentence (`\emph{cost}`) will not
+match, and the finding will quote exactly what it looked for.
+
+## The row ids, and what each one runs
+
+The `id` cell is the **join key** — the same string appears in this template, in every paper's
+scorecard, in `Requires` cells, in `../scripts/pipeline-edges.mjs` and in the skills' own prose. So
+the scorecard explains itself where it lives:
+
+| id            | section    | what it is                                                                | run by                          |
+| ------------- | ---------- | ------------------------------------------------------------------------- | ------------------------------- |
+| `idea`        | SETUP      | is the idea worth a paper?                                                | `research-ideate`               |
+| `priorwork`   | SETUP      | full competitor sweep, before drafting                                    | `map-prior-work`                |
+| `venue`       | SETUP      | pick where it goes                                                        | `find-venue`                    |
+| `venuebar`    | SETUP      | what this venue rewards; strengthening levers                             | `study-accepted-papers`         |
+| `access`      | SETUP      | 🔴 _can you physically upload?_ — account, ACTIVE profile, portal, form   | `plan-paper-timeline`           |
+| `schedule`    | SETUP      | the backwards plan and its calendar events                                | `plan-paper-timeline`           |
+| `frame`       | SETUP      | the claim, stated **before** the runs                                     | `argument-arc` (frame mode)     |
+| `study`       | LOOP       | the empirical study + reproduction artifact                               | `build-benchmark`               |
+| `draft`       | LOOP       | the paper itself                                                          | `draft-paper`                   |
+| `arc`         | LOOP       | does the argument carry end to end?                                       | `argument-arc`                  |
+| `cites`       | CONTINUOUS | every `\cite` is real, and the delta over the nearest neighbour is stated | `verify-citations`              |
+| `siblings`    | CONTINUOUS | deep read of a competing/concurrent paper                                 | `analyze-sibling-paper`         |
+| `priordelta`  | CONTINUOUS | re-sweep restricted to work posted since                                  | `map-prior-work` (delta)        |
+| `render`      | CONTINUOUS | build the PDF; the page count everything else is blocked on               | `render-paper`                  |
+| `numbers`     | CONTINUOUS | every printed quantity traces to committed data                           | the paper's own numbers gate    |
+| `structure`   | GATES      | the developmental/editorial verdict + cut plan                            | `tighten-paper`                 |
+| `writing`     | GATES      | writing craft + the persona stall inventory                               | `grade-paper-writing`           |
+| `panel`       | GATES      | the simulated PC decision                                                 | `pc-panel-review`               |
+| `artifact`    | GATES      | defects found by **executing** the artifact                               | `pc-panel-review` §C            |
+| `claims`      | GATES      | claim-preservation diff after every rewrite                               | Fable diff (`writing-craft.md`) |
+| `harden`      | GATES      | the multi-axis pre-submit gate                                            | `harden-paper`                  |
+| `submit`      | GATES      | the upload itself                                                         | `submit-paper`                  |
+| `coldread`    | GATES      | a reader with no context on the changed prose                             | `cold-read-diff`                |
+| `cameraready` | AFTER      | de-anonymize, disclose, archive                                           | `camera-ready`                  |
+| `extend`      | AFTER      | the ≥30%-new follow-on at a stronger venue                                | `extend-paper`                  |
+
+🔴 **Ids are WORDS, and that is load-bearing** (renamed 2026-08-09 from two-letter codes). Two
+letters had run out and had already collided twice, both live in this repo: `Cr` was
+`cold-read-diff` in one table of `compile-rules-2026` and `Camera-ready` in another — one join key,
+two rows, merged by every consumer — and `Id` (research-ideate) was the header word `id`, so the
+parser's header-row filter silently ate that row on **every** scorecard for as long as the template
+existed. Neither is expressible now. Keep new ids lowercase, one word, and say what the row is.
+
+### How `pipeline-check.mjs` reads this file (rewritten 2026-08-09 to parse markdown, not match it)
+
+Both defects above, plus a third — two rows sharing an id merging in silence — had one cause: a
+regular expression guessing at document structure. The parser is now an AST walk (`markdown-it`),
+and two rules decide what it reads. They are stated here because the checker's findings point at
+them:
+
+- **A heading names a section iff its FIRST Latin-script word is `SETUP` / `LOOP` / `CONTINUOUS` /
+  `GATES` / `AFTER`.** Anything after that word is free: `### GATES (re-run 06.08)` is GATES. A
+  heading with a different first word — or none, like `### Гейты, прогнанные 06.08` — names no
+  section and CLOSES the previous one. Headings are **not** translated or fuzzy-matched: guessing
+  what a heading means is the defect, not the fix. Rows in such a region are reported
+  (`unattributed-row`), never dropped.
+- **A table is part of the scorecard iff its header row's first cell is `id`** (also `код`, `code`).
+  The header comes from the table's real `thead`, so there is nothing to guess and a data row
+  reading `| Id |` is a row. Prose tables in a scorecard — the panel must-fix table, a ceiling
+  classification — have no id column and are correctly ignored. The pre-2026-08 single-table format
+  keys its first column `#`, a row number; it does not parse, and the checker says so by name.
+
+**A row is never dropped in silence.** Unattributable, unreadable and duplicate ids are findings.
+
+## Rules
+
+- **One row per unit of work; update the row in the SAME commit as the work.** Status + ISO date +
+  result + open findings + a link to the durable artifact (`siblings/`, a saved grader report).
+- **Four sections, because the four kinds fail differently** (see `paper-pipeline` → "There is no
+  stage 7"). SETUP is done-or-not. LOOP iterates. **CONTINUOUS is never done** — its date is compared
+  against the paper's own mtime, and a `☑` older than the text is a _stale pass_, which reads as
+  green and is not. GATES are blocked on named inputs.
+- 🔴 **The `Requires` cells above are not free-form — do not thin them.** The canonical edge set lives
+  in `../scripts/pipeline-edges.mjs`, with the sentence from each skill's own SKILL.md that admitted
+  the edge, and the ESLint rule `pipeline/undeclared-input` (`eslint-rules/pipeline-status.mjs`;
+  until 2026-08-26 `pipeline-check.mjs`) reports any cell that drops one. The cell
+  was retyped by hand for every paper until 2026-08-08, and `compile-rules-2026` had lost four of
+  them — including `arc`, which argument-arc's SKILL.md says must precede `tighten-paper`. A dropped
+  edge is invisible: `gate-missing-input` can only enforce what the cell says. Adding an input the
+  table does not know about is fine and reports nothing.
+- 🔴 **Row `access` is the one that is not about the paper, and it is the one that bites.** _Can you
+  physically upload?_ — account exists, profile is **ACTIVE** rather than "pending moderation", the
+  portal is reachable, you know the form's fields and its artifact-hosting requirement. On
+  `compile-rules-2026` this was the sole critical-path item at T−4 days: the paper was finished and
+  there was nowhere to put it, because OpenReview requires an active profile to submit and moderation
+  runs **up to two weeks**. Turn it green in the first week, not the last. Record the date it went
+  ACTIVE, not just the tick.
+- **Row `frame` costs one paragraph and saves a study.** State what the paper will claim _before_ the
+  runs. Reframing after the data is collected is how experiments get thrown away — observed on
+  `compile-rules-2026`, and the reason `argument-arc` has a rebuild mode at all.
+- **Status legend:** `☑` done/pass · `◐` partial (say what's missing) · `☐` not run · `⚠` ran, findings
+  open · `n/a` not applicable to this paper/venue.
+- **Never inflate.** A unit that didn't run _this_ cycle is `☐` — do not borrow credit from a prior
+  submission you can't point to. Approximate or from-memory grades get `≈`.
+- **The verdict line leads WORST-GATE-FIRST.** One sentence: submit-ready or not, and the single thing
+  blocking. "Ready" is a vector (science/defects, readability, structure, claims-honest, mechanical) —
+  **any failing gate caps the verdict**. **A high accept-probability alone is never "submit-ready"**:
+  an ~88% panel number once masked failing readability and structure on a paper a human found
+  exhausting, and the checker now refuses a verdict line that is only a percentage.
+- **Distinct from `SUBMIT-CHECKLIST.md`** (venue-format compliance: page limit, blind model, portal).
+  This tracks _quality completeness_. Both green before upload.
+
+---
+
+The template itself is a FILE, not this page: `templates/paper/PIPELINE-STATUS.md` at the package
+root (beside `templates/paper/paper.tex` and `paper.md`). `paperlint new <name>` copies it with the paper's
+name filled in, and a project that keeps its own richer scorecard puts one at
+`<papers>/.template/PIPELINE-STATUS.md`, which `paperlint new` prefers. It used to live in a fenced block
+here, which meant a command would have had to parse markdown to extract it, and a human had to copy
+it by hand. The copy carries `researchQuestion` (empty) and no `stages` — a new paper has shipped
+nothing.
+
+---
+
+## How to use
+
+- **"Is paper X ready / what's checked?"** → read `<paper-dir>/PIPELINE-STATUS.md`. Report the verdict
+  line + any `☐`/`⚠`/`◐` rows. Do not re-derive from chat.
+- **After running anything** (here or in a subagent) → update its row before moving on; commit with the
+  work.
+- **Before submit** → every non-`n/a` row is `☑`, or a consciously-accepted `◐`/`⚠` named in the verdict
+  line. **An unrun GATE is not a valid "anti-churn" skip if capacity exists: run it.** Anti-churn covers
+  re-polishing prose that already passed a gate, never skipping a gate that never ran. And a **stale
+  CONTINUOUS row is worse than an empty one** — it reads as green while being false.

package/skills/paper-pipeline/references/review-ratchet.md

@@ -0,0 +1,36 @@
+# The review ratchet — how to fix a caught overclaim, and how to pay for it
+
+Cited by `paper-adversarial-review`, `pc-panel-review` and `harden-paper`. It lived as three
+near-identical paragraphs in those three skills until 2026-07-30; three copies of a rule drift, and
+the drift is invisible because each copy reads fine on its own.
+
+## The rule
+
+**Review only ever ADDS text.** Every round deposits hedges, caveats, clarifications and one more
+citation. Nothing in a review round removes anything — that is what makes it a ratchet, and why a
+paper that survived four rounds reads like armor.
+
+**When a review catches an overclaim, inline-hedging is the LAST resort.** In order:
+
+1. **TIGHTEN** the claim to what is actually true. Usually shorter _and_ correct — the strongest fix
+   and the one that gets skipped, because it means admitting the smaller result is the result.
+2. **CUT it.** A claim you cannot support is not costing you what you think it is.
+3. **Move the caveat to Threats to Validity**, stated once. Adversarial reviewers read Threats first,
+   and an honest limitation there reads as strength; the same sentence inline reads as flinching.
+4. **Only then inline-hedge** — and if you find yourself here often, the contribution is mis-sized,
+   not under-qualified.
+
+## Paying the ratchet back
+
+**After applying a round's consensus fixes, run `tighten-paper`** to strip the mass the round
+deposited. Structure-pruning is the counterweight; without it the paper monotonically inflates.
+
+Observed cost of skipping it: one session, seven overclaim fixes plus one panel round added roughly a
+page of pure hedge density.
+
+## What this rule does NOT license
+
+- It is not permission to skip a gate. Trimming what a _passed_ gate deposited is anti-churn; not
+  running a gate at all is a hole.
+- It is not a reason to cut a **Limitations** bullet. Those are the hedges that earned their place —
+  they are stated once, in the right section, which is exactly what step 3 asks for.