paperlint 2.0.0

package/skills/paper-pipeline/scripts/extract-ref-facts.mutations.mjs

@@ -0,0 +1,54 @@
+/**
+ * extract-ref-facts.mutations.mjs — plant the defect `extract-ref-facts.harness.mjs` claims to catch, and require it to die.
+ *
+ * Run: `node skills/paper-pipeline/scripts/extract-ref-facts.mutations.mjs` (it rewrites the source and restores
+ * it, so it is not part of the plain test run).
+ *
+ * WHY IT EXISTS. A harness whose success state is silence cannot be observed working: «it passed»
+ * and «it cannot fail» produce identical output. Written 2026-09-12, when the mechanism moved into
+ * this package and `scripts/run-mutations.mjs` reported this harness as one nothing could kill.
+ *
+ * ⚠️ HONEST SCOPE: ONE case. It kills the checker outright — the shape that makes every other
+ * assertion in the harness vacuous at once — and that is the property worth holding first, not the
+ * only one worth holding. Narrower cases were probed and left out DELIBERATELY: the assertions
+ * they trip carry no message of their own (a bare deepEqual diff), so `expect` could only match
+ * text any unrelated failure also prints, and a case killed by someone else's assertion is not
+ * evidence about its own defect. Adding one means giving that assertion a message first.
+ */
+import { dirname, join } from "node:path";
+import { fileURLToPath } from "node:url";
+import { runMutations } from "../../../lib/mutation-driver.mjs";
+import { consumerRoot } from "./consumer.mjs";
+
+const HERE = dirname(fileURLToPath(import.meta.url));
+const SRC = join(HERE, "extract-ref-facts.mjs");
+const HARNESS = join(HERE, "extract-ref-facts.harness.mjs");
+
+const MUTATIONS = [
+  [
+    "the markdown reference list is never parsed",
+    [
+      [
+        SRC,
+        "export function parseMarkdownRefs(text) {",
+        "export function parseMarkdownRefs(text) {\n  if (1) return [];",
+      ],
+    ],
+    "heading must close the reference list",
+    "the markdown leg entirely — a paper whose bibliography is prose yields zero refs and no complaint",
+  ],
+];
+
+process.exit(
+  runMutations({
+    root: consumerRoot(),
+    runner: "vigiles",
+    cases: MUTATIONS.map(([name, edits, expect, disables]) => ({
+      name,
+      disables,
+      edits,
+      harness: HARNESS,
+      expect,
+    })),
+  }),
+);

package/skills/paper-pipeline/scripts/fixtures/clean/PIPELINE-STATUS.md

@@ -0,0 +1,51 @@
+# PIPELINE-STATUS — fixture: nothing to report
+Venue: Fixture Workshop · Deadline: 2027-12-31 · Blind: double · State: drafting
+Updated: 2026-08-03
+
+**Readiness verdict:** Submit-ready — every gate green, artifact reproduces clean, nothing blocking.
+
+This fixture deliberately carries NO paper source file, so `newestSourceDate` is null and the
+staleness comparison has nothing to compare against. That keeps the clean case deterministic
+regardless of when the repo was checked out — the staleness rule is exercised by `dirty/`.
+
+### SETUP
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| idea | Validate idea | research-ideate | ☑ | 2026-08-01 | go. judge: opus, judge: author | — |
+| priorwork | Map competitors | map-prior-work | ☑ | 2026-08-01 | 6 siblings. judge: sonnet, judge: author | siblings/ |
+| venue | Pick venue | find-venue | ☑ | 2026-08-01 | fixture. judge: sonnet, judge: author | — |
+| venuebar | Venue bar / levers | study-accepted-papers | ☑ | 2026-08-01 | 34 measured. judge: sonnet, judge: author | — |
+| **access** | 🔴 **Can you physically submit?** | plan-paper-timeline | ☑ | 2026-08-01 | profile ACTIVE since 2026-08-01. judge: author, judge: author | — |
+| schedule | Schedule + calendar | plan-paper-timeline | ☑ | 2026-08-01 | —. judge: author, judge: author | — |
+| **frame** | **Claim stated before the runs** | argument-arc (frame) | ☑ | 2026-07-20 | one paragraph filed | — |
+
+### LOOP
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| study | Study + artifact | build-benchmark | ☑ | 2026-08-01 | ran | — |
+| draft | Draft | draft-paper | ☑ | 2026-08-01 | — | — |
+| arc | Argument arc | argument-arc | ☑ | 2026-08-01 | holds. judge: opus, judge: author | — |
+
+### CONTINUOUS
+| id | Trigger | Skill | Status | Date | Result | Open |
+|----|---------|-------|--------|------|--------|------|
+| cites | any cite added/moved | verify-citations | ☑ | 2026-08-02 | clean | — |
+| siblings | a sibling surfaces | analyze-sibling-paper | ☑ | 2026-08-02 | 6 read. judge: opus, judge: author | siblings/ |
+| priordelta | framing moved | map-prior-work (delta) | ☑ | 2026-08-02 | no new | — |
+| render | any source edit | render-paper + page count | ☑ | 2026-08-02 | 8 pages / limit 8 | — |
+
+### GATES
+| id | Gate | Skill | Requires | Status | Date | Result | Open |
+|----|------|-------|----------|--------|------|--------|------|
+| structure | Structure | tighten-paper | render, arc | ☑ | 2026-08-02 | cut plan applied. judge: opus, judge: author | — |
+| writing | Writing craft | grade-paper-writing | draft, arc, structure | ☑ | 2026-08-02 | 44/60. judge: sonnet, judge: author | — |
+| panel | Review decision | pc-panel-review | structure, writing | ☑ | 2026-08-03 | accept; judge: opus, judge: author | — |
+| claims | Claim preservation | Fable diff | draft, writing | ☑ | 2026-08-03 | clean | — |
+| harden | Harden (all axes) | harden-paper | panel, structure, writing, cites, priordelta | ☑ | 2026-08-03 | all axes green ; judge: opus, judge: author | — |
+| submit | Submit | submit-paper | harden, access | ☑ | 2026-08-03 | uploaded | — |
+
+### AFTER
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| cameraready | Camera-ready | camera-ready | n/a | — | pre-acceptance | — |
+| extend | Extend | extend-paper | n/a | — | post-acceptance | — |

package/skills/paper-pipeline/scripts/fixtures/dirty/PIPELINE-STATUS.md

@@ -0,0 +1,52 @@
+# PIPELINE-STATUS — fixture: every failure this checker exists for
+Venue: Fixture Workshop · Deadline: 2026-08-06 · Blind: double · State: drafting
+Updated: 2026-01-01
+
+**Readiness verdict:** 88% accept probability, looking good.
+
+The `structure` row is missing `arc` from its Requires cell **on purpose** — that is the planted
+`undeclared-input`, copied from a defect found live on a real paper in August 2026. Every
+other gate cell here is the canonical set from `../../pipeline-edges.mjs`, so exactly one edge is
+undeclared and the finding cannot come from somewhere else.
+
+### SETUP
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| idea | Validate idea | research-ideate | ☑ | 2026-01-01 | go | — |
+| priorwork | Map competitors | map-prior-work | ☑ | 2026-01-01 | 6 siblings | siblings/ |
+| venue | Pick venue | find-venue | ☑ | 2026-01-01 | fixture | — |
+| venuebar | Venue bar / levers | study-accepted-papers | ☐ | — | — | — |
+| **access** | 🔴 **Can you physically submit?** | plan-paper-timeline | ☐ | — | profile pending moderation | — |
+| schedule | Schedule + calendar | plan-paper-timeline | ☑ | 2026-01-01 | — | — |
+| **frame** | **Claim stated before the runs** | argument-arc (frame) | ☐ | — | — | — |
+
+### LOOP
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| study | Study + artifact | build-benchmark | ☑ | 2026-01-01 | ran | — |
+| draft | Draft | draft-paper | ☑ | 2026-01-01 | — | — |
+| arc | Argument arc | argument-arc | ☐ | — | — | — |
+
+### CONTINUOUS
+| id | Trigger | Skill | Status | Date | Result | Open |
+|----|---------|-------|--------|------|--------|------|
+| cites | any cite added/moved | verify-citations | ☑ | 2026-01-01 | clean | — |
+| siblings | a sibling surfaces | analyze-sibling-paper | ☑ | — | done | siblings/ |
+| priordelta | framing moved | map-prior-work (delta) | ☐ | — | — | — |
+| render | any source edit | render-paper + page count | ☑ | 2026-01-01 | 8 pages / limit 8 | — |
+
+### GATES
+| id | Gate | Skill | Requires | Status | Date | Result | Open |
+|----|------|-------|----------|--------|------|--------|------|
+| structure | Structure | tighten-paper | render | ☑ | 2026-01-01 | cut plan | — |
+| writing | Writing craft | grade-paper-writing | draft, arc, structure | ☐ | — | — | — |
+| panel | Review decision | pc-panel-review | structure, writing | ☑ | 2026-01-01 | borderline | — |
+| claims | Claim preservation | Fable diff | draft, writing | ☐ | — | — | — |
+| harden | Harden (all axes) | harden-paper | panel, structure, writing, cites, priordelta | ☐ | — | — | — |
+| submit | Submit | submit-paper | harden, access | ☐ | — | — | — |
+
+### AFTER
+| id | Work | Skill | Status | Date | Result | Open |
+|----|------|-------|--------|------|--------|------|
+| cameraready | Camera-ready | camera-ready | n/a | — | pre-acceptance | — |
+| extend | Extend | extend-paper | n/a | — | post-acceptance | — |

package/skills/paper-pipeline/scripts/fixtures/dirty/paper.md

@@ -0,0 +1,6 @@
+# Fixture paper source
+
+This file exists only so `newestSourceDate` finds something. Its mtime is whenever the
+repo was checked out, which is always later than the 2026-01-01 dates in the fixture's
+CONTINUOUS rows — so the `stale-continuous` check fires deterministically without the
+test having to touch mtimes.

package/skills/paper-pipeline/scripts/fixtures/real-bib/refs.bib

@@ -0,0 +1,153 @@
+% real-world-refs.bib — НАСТОЯЩИЙ BibTeX, а не написанный руками. Заморожен 12.09.2026.
+%
+% 🔴 ПОЧЕМУ ФИКСТУРА НАСТОЯЩАЯ. Прежний парсер падал на форме `}}` в конце последнего поля, и
+% комментарий в `extract-ref-facts.harness.mjs` прямо говорит: фикстура, написанная руками, почти
+% наверняка закрыла бы запись на своей строке — то есть тест прошёл бы, а корпус нет. Поэтому
+% здесь лежат подлинные байты (27 записей, 28 вхождений `}}`), а не их аккуратная имитация.
+%
+% ОТКУДА. Библиография ПРИНЯТОЙ статьи (труды выходят 12.10.2026), то есть список чужих
+% опубликованных работ, который и так станет публичным. Проверено перед копированием: ноль
+% самоцитирований, ноль упоминаний автора, ноль приватных путей — только цитируемые чужие работы.
+% Статья под анонимным ревью для этой цели НЕ бралась намеренно: набор ссылок сам по себе
+% отпечаток, и публиковать его до раскрытия нельзя.
+%
+% ЧТО ЭТИМ ПРОВЕРЯЕТСЯ: что разбор переживает настоящие формы. Число записей считается ИЗ ФАЙЛА
+% (`^@`), а не вписано константой — константа устареет при первой дописанной записи.
+@inproceedings{jimenez2024swebench,
+  author={Jimenez, Carlos E. and Yang, John and Wettig, Alexander and Yao, Shunyu and Pei, Kexin and Press, Ofir and Narasimhan, Karthik},
+  title={{SWE-bench}: Can Language Models Resolve Real-World {GitHub} Issues?},
+  booktitle={International Conference on Learning Representations (ICLR)}, year={2024},
+  doi={10.48550/arXiv.2310.06770}}
+@inproceedings{yang2024sweagent,
+  author={Yang, John and Jimenez, Carlos E. and Wettig, Alexander and Lieret, Kilian and Yao, Shunyu and Narasimhan, Karthik and Press, Ofir},
+  title={{SWE-agent}: Agent-Computer Interfaces Enable Automated Software Engineering},
+  booktitle={Advances in Neural Information Processing Systems (NeurIPS)}, year={2024},
+  doi={10.48550/arXiv.2405.15793}}
+@inproceedings{jain2025livecodebench,
+  author={Jain, Naman and Han, King and Gu, Alex and Li, Wen-Ding and Yan, Fanjia and Zhang, Tianjun and Wang, Sida and Solar-Lezama, Armando and Sen, Koushik and Stoica, Ion},
+  title={{LiveCodeBench}: Holistic and Contamination Free Evaluation of Large Language Models for Code},
+  booktitle={International Conference on Learning Representations (ICLR)}, year={2025},
+  doi={10.48550/arXiv.2403.07974}}
+@article{chen2021humaneval,
+  author={Chen, Mark and Tworek, Jerry and Jun, Heewoo and Yuan, Qiming and others},
+  title={Evaluating Large Language Models Trained on Code}, journal={arXiv preprint arXiv:2107.03374}, year={2021},
+  doi={10.48550/arXiv.2107.03374}}
+@inproceedings{jiang2023llmlingua,
+  author={Jiang, Huiqiang and Wu, Qianhui and Lin, Chin-Yew and Yang, Yuqing and Qiu, Lili},
+  title={{LLMLingua}: Compressing Prompts for Accelerated Inference of Large Language Models},
+  booktitle={Conference on Empirical Methods in Natural Language Processing (EMNLP)}, year={2023}, note={arXiv:2310.05736},
+  doi={10.18653/v1/2023.emnlp-main.825}}
+@inproceedings{zheng2024sglang,
+  author={Zheng, Lianmin and Yin, Liangsheng and Xie, Zhiqiang and others},
+  title={{SGLang}: Efficient Execution of Structured Language Model Programs},
+  booktitle={Advances in Neural Information Processing Systems (NeurIPS)}, year={2024},
+  doi={10.48550/arXiv.2312.07104}}
+@article{chen2023frugalgpt,
+  author={Chen, Lingjiao and Zaharia, Matei and Zou, James},
+  title={{FrugalGPT}: How to Use Large Language Models While Reducing Cost and Improving Performance},
+  journal={arXiv preprint arXiv:2305.05176}, year={2023}, doi={10.48550/arXiv.2305.05176}}
+@article{schwartz2020greenai,
+  author={Schwartz, Roy and Dodge, Jesse and Smith, Noah A. and Etzioni, Oren},
+  title={Green {AI}}, journal={Communications of the ACM}, volume={63}, number={12}, pages={54--63}, year={2020},
+  doi={10.1145/3381831}}
+% 2026-08-24, found by scripts/bib-authors.mjs (the THIRD entry with this same defect): the
+% order below was the arXiv preprint's. NeurIPS D&B 2021 orders them Raji, Denton, Bender,
+% Hanna, Paullada. We cite the proceedings. Confirmed on the DBLP API (which returns the
+% NeurIPS and CoRR records separately) and against arXiv 2111.15366.
+@inproceedings{raji2021benchmark,
+  author={Raji, Inioluwa Deborah and Denton, Emily and Bender, Emily M. and Hanna, Alex and Paullada, Amandalynne},
+  title={{AI} and the Everything in the Whole Wide World Benchmark},
+  booktitle={NeurIPS Datasets and Benchmarks Track}, year={2021},
+  doi={10.48550/arXiv.2111.15366}}
+@article{biderman2024lessons,
+  author={Biderman, Stella and Schoelkopf, Hailey and Sutawika, Lintang and Gao, Leo and Tow, Jonathan and others},
+  title={Lessons from the Trenches on Reproducible Evaluation of Language Models},
+  journal={arXiv preprint arXiv:2405.14782}, year={2024}, doi={10.48550/arXiv.2405.14782}}
+@inproceedings{jacobs2021measurement,
+  author={Jacobs, Abigail Z. and Wallach, Hanna},
+  title={Measurement and Fairness},
+  booktitle={ACM Conference on Fairness, Accountability, and Transparency (FAccT)}, pages={375--385}, year={2021},
+  doi={10.1145/3442188.3445901}}
+@inproceedings{yao2023react,
+  author={Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik and Cao, Yuan},
+  title={{ReAct}: Synergizing Reasoning and Acting in Language Models},
+  booktitle={International Conference on Learning Representations (ICLR)}, year={2023},
+  doi={10.48550/arXiv.2210.03629}}
+% 2026-08-24: Eric Hambro was MISSING below. The NeurIPS 2023 version has 9 authors, the arXiv
+% preprint 8 — we cite the proceedings, so the proceedings author list is the correct one.
+% Confirmed on the DBLP API, which lists both records side by side (NeurIPS 2023 vs CoRR 2023).
+% NB: this comment lives OUTSIDE the entry on purpose — `%` is NOT a comment inside a BibTeX
+% entry, and putting it there made BibTeX skip the whole record silently (measured the same day).
+@inproceedings{schick2023toolformer,
+  author={Schick, Timo and Dwivedi-Yu, Jane and Dess{\`i}, Roberto and Raileanu, Roberta and Lomeli, Maria and Hambro, Eric and Zettlemoyer, Luke and Cancedda, Nicola and Scialom, Thomas},
+  title={{Toolformer}: Language Models Can Teach Themselves to Use Tools},
+  booktitle={Advances in Neural Information Processing Systems (NeurIPS)}, year={2023},
+  doi={10.48550/arXiv.2302.04761}}
+@article{liu2024seagentsurvey,
+  author={Liu, Junwei and Wang, Kaixin and Chen, Yixuan and Peng, Xin and Chen, Zhenpeng and Zhang, Lingming and Lou, Yiling},
+  title={Large Language Model-Based Agents for Software Engineering: A Survey},
+  journal={arXiv preprint arXiv:2409.02977}, year={2024}, doi={10.48550/arXiv.2409.02977}}
+@inproceedings{agarwal2021precipice,
+  author={Agarwal, Rishabh and Schwarzer, Max and Castro, Pablo Samuel and Courville, Aaron and Bellemare, Marc G.},
+  title={Deep Reinforcement Learning at the Edge of the Statistical Precipice},
+  booktitle={Advances in Neural Information Processing Systems (NeurIPS)}, year={2021},
+  doi={10.48550/arXiv.2108.13264}}
+@article{liang2023helm,
+  author={Liang, Percy and Bommasani, Rishi and Lee, Tony and others},
+  title={Holistic Evaluation of Language Models}, journal={Transactions on Machine Learning Research (TMLR)}, year={2023},
+  doi={10.48550/arXiv.2211.09110}}
+@article{kapoor2024agentsmatter,
+  author={Kapoor, Sayash and Stroebl, Benedikt and Siegel, Zachary S. and Nadgir, Nitya and Narayanan, Arvind},
+  title={{AI} Agents That Matter}, journal={arXiv preprint arXiv:2407.01502}, year={2024},
+  doi={10.48550/arXiv.2407.01502}}
+% 2026-08-24: order below was the PREPRINT's (Tay last). ICLR 2022 puts Tay second. We cite ICLR.
+% ⚠️ PIPELINE-STATUS recorded "dehghani author-order fixed" on 2026-07-22 — it was fixed to the
+% arXiv order, which is a different list from the published one. Confirmed on the DBLP API.
+@inproceedings{dehghani2022efficiency,
+  author={Dehghani, Mostafa and Tay, Yi and Arnab, Anurag and Beyer, Lucas and Vaswani, Ashish},
+  title={The Efficiency Misnomer},
+  booktitle={International Conference on Learning Representations (ICLR)}, year={2022},
+  doi={10.48550/arXiv.2110.12894}}
+@article{barke2023grounded,
+  author={Barke, Shraddha and James, Michael B. and Polikarpova, Nadia},
+  title={Grounded {Copilot}: How Programmers Interact with Code-Generating Models},
+  journal={Proceedings of the ACM on Programming Languages (OOPSLA)}, volume={7}, pages={85--111}, year={2023},
+  doi={10.1145/3586030}}
+@article{bai2026spend,
+  author={Bai, Longju and Huang, Zhemin and Wang, Xingyao and Sun, Jiao and Mihalcea, Rada and Brynjolfsson, Erik and Pentland, Alex and Pei, Jiaxin},
+  title={How Do {AI} Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding Tasks},
+  journal={arXiv preprint arXiv:2604.22750}, year={2026}, doi={10.48550/arXiv.2604.22750}}
+@misc{jetbrains2026caveman,
+  author={Shiryaev, Denis},
+  title={Does Speaking to Agents Like Cavemen Really Save 65\% of Tokens? {We} Test},
+  year={2026},
+  howpublished={{JetBrains} {AI} blog},
+  url={https://blog.jetbrains.com/ai/2026/07/speak-to-ai-agents-like-cavemen-tosave-tokens/},
+  note={Independent A/B test of the caveman skill on 86 {Claude Code} tasks; about 8.5\% output savings vs.\ the advertised 65\%. Secondary coverage: \url{https://www.infoworld.com/article/4193775/talk-like-a-caveman-prompts-save-tokens-but-far-less-than-promised.html}}}
+@inproceedings{lago2024threats,
+  author={Lago, Patricia and Runeson, Per and Song, Qunying and Verdecchia, Roberto},
+  title={Threats to Validity in Software Engineering -- hypocritical paper section or essential analysis?},
+  booktitle={Proceedings of the 18th ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM)},
+  pages={314--324}, year={2024}, doi={10.1145/3674805.3686691}}
+@article{erol2025costofpass,
+  author={Erol, Mehmet Hamza and El, Batu and Suzgun, Mirac and Yuksekgonul, Mert and Zou, James},
+  title={Cost-of-Pass: An Economic Framework for Evaluating Language Models},
+  journal={arXiv preprint arXiv:2504.13359}, year={2025}, doi={10.48550/arXiv.2504.13359}}
+@inproceedings{bean2025construct,
+  author={Bean, Andrew M. and Kearns, Ryan Othniel and Romanou, Angelika and others},
+  title={Measuring What Matters: Construct Validity in Large Language Model Benchmarks},
+  booktitle={NeurIPS Datasets and Benchmarks Track}, year={2025},
+  doi={10.48550/arXiv.2511.04703}}
+@inproceedings{liu2023evalplus,
+  author={Liu, Jiawei and Xia, Chunqiu Steven and Wang, Yuyao and Zhang, Lingming},
+  title={Is Your Code Generated by {ChatGPT} Really Correct? Rigorous Evaluation of Large Language Models for Code Generation},
+  booktitle={Advances in Neural Information Processing Systems (NeurIPS)}, year={2023},
+  doi={10.48550/arXiv.2305.01210}}
+@article{liang2025swebenchillusion,
+  author={Liang, Shanchao and Garg, Spandan and Zilouchian Moghaddam, Roshanak},
+  title={The {SWE-Bench} Illusion: When State-of-the-Art {LLMs} Remember Instead of Reason},
+  journal={arXiv preprint arXiv:2506.12286}, year={2025}, doi={10.48550/arXiv.2506.12286}}
+@article{yao2024taubench,
+  author={Yao, Shunyu and Shinn, Noah and Razavi, Pedram and Narasimhan, Karthik},
+  title={{$\tau$-bench}: A Benchmark for Tool-Agent-User Interaction in Real-World Domains},
+  journal={arXiv preprint arXiv:2406.12045}, year={2024}, doi={10.48550/arXiv.2406.12045}}