cd /news/ai-agents/academic-writing-kit-for-claude-code · home topics ai-agents article
[ARTICLE · art-42537] src=github.com ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

Academic-writing kit for Claude Code

Josef Šlerka released an open-source academic-writing kit for Claude Code and Cursor that guides an AI agent through producing a scholarly paper from raw insight to finished study while keeping the human as the author. The kit, built on Andrej Karpathy's LLM-wiki pattern, is available in Czech and English and includes immutable raw sources, model-owned artifacts, and a hostile-review gate. It is licensed under MIT and designed to ensure human accountability for truth.

read2 min views1 publishedJun 28, 2026
Academic-writing kit for Claude Code
Image: source

A reproducible pipeline that has an AI agent (Claude Code or Cursor) walk you through writing a scholarly paper — from a raw insight to a finished, source-grounded study — while a human stays the author and the one who vouches for the truth. Built on Andrej Karpathy's LLM-wiki pattern (immutable raw sources → model-owned artifacts, grounding, verbatim provenance, a hostile-review gate).

This repo ships the kit in two languages, as two self-contained directories:

cs/ | 🇨🇿 Czech — the original, empirically tuned on a study of Karel Čapek | en/ | 🇬🇧 English — a translation of the same kit |

Pick a directory, cd

into it, and follow its README

. Each directory is a complete project: its own CLAUDE.md

(the brain), slash commands, and folders.

The Czech kit is the one that was run, broken, and fixed many times — its rules are scars from real failures. The English kit is a faithful translation of that same structure; the architecture and rules transfer, but it hasn't been put through the same empirical loop, so re-tune it by running it on a real paper and adding a rule whenever the model slips.

Don't commit your corpus. raw_primary/

and raw_secondary/

are git-ignored in both kits: primary texts and especially secondary literature are usually under copyright. Keep them local.

Reprodukovatelný pipeline, ve kterém vás AI agent (Claude Code nebo Cursor) provede psaním odborné studie — od syrového nápadu k hotové, ve zdrojích ukotvené práci — přičemž autorem a tím, kdo ručí za pravdivost, zůstává člověk. Postaveno na LLM-wiki vzoru Andreje Karpathyho (neměnné syrové zdroje → artefakty vlastněné modelem, ukotvení, doslovná provenience, brána nepřátelské oponentury).

Tento repozitář dodává kit ve dvou jazycích, jako dva samostatné adresáře:

cs/ | 🇨🇿 Čeština — originál, empiricky vyladěný na studii o Karlu Čapkovi | en/ | 🇬🇧 Angličtina — překlad téhož kitu |

Vyberte si adresář, cd

do něj a řiďte se jeho README

. Každý adresář je kompletní projekt: má vlastní CLAUDE.md

(mozek), slash příkazy a složky. Český kit je ten, který byl mnohokrát spuštěn, rozbit a opraven — jeho pravidla jsou jizvy ze skutečných selhání. Anglický kit je věrný překlad téže struktury; architektura a pravidla se přenášejí, ale neprošel stejnou empirickou smyčkou, takže ho doladěte tím, že ho spustíte na reálné práci a přidáte pravidlo pokaždé, když model uklouzne.

Nekomitujte svůj korpus. raw_primary/

a raw_secondary/

jsou v obou kitech v gitignore: primární texty a zvláště sekundární literatura bývají chráněny autorským právem. Nechte si je lokálně.

Workflow: Josef Šlerka, "AI-Assisted Academic Writing Pipeline" (spring 2026). Adapt freely. / Adaptujte volně.

License / Licence: MIT © 2026 Josef Šlerka.

── more in #ai-agents 4 stories · sorted by recency
── more on @josef šlerka 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/academic-writing-kit…] indexed:0 read:2min 2026-06-28 ·