What is paper-to-code?

A workflow for turning an academic paper, such as one on arXiv, into code that can be checked. It records unclear details and ties implementation decisions to the paper or clearly labels them as choices.

Use it to implement, reproduce, or prototype a method described in a quantitative research paper and run sanity checks on the result.

It prevents missing details from being silently filled in with assumptions that may not match the research.

paper-to-code — Skill for Claude Code by AlexShchuka/neuro-matrix

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add AlexShchuka/neuro-matrix --skill paper-to-code

Clone the repo

git clone --depth 1 https://github.com/AlexShchuka/neuro-matrix

Made for: Claude Code.

Or install neuro-matrix, the plugin that ships this one along with the rest of its 6 skills, 5 agents, 4 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/alexshchuka/neuro-matrix/paper-to-code/github.svg)](https://agentmods.dev/skills/alexshchuka/neuro-matrix/paper-to-code)

Your own site

<a href="https://agentmods.dev/skills/alexshchuka/neuro-matrix/paper-to-code"><img src="https://agentmods.dev/badge/skills/alexshchuka/neuro-matrix/paper-to-code/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/alexshchuka/neuro-matrix/paper-to-code"><img src="https://agentmods.dev/badge/skills/alexshchuka/neuro-matrix/paper-to-code.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 65 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 857 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00065	$0.00857
Opus 5.5	$0.00026	$0.00343
Sonnet 5.5	$0.00013	$0.00171
Haiku 4.5	$0.00006	$0.00086

Measured 2mo ago against content hash 446d2821ae7d, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

paper-to-code scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/paper-to-code/SKILL.md · 52 lines

How it starts

The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Paper-to-code: implementation without invention

BLUF: the paper is the only source of truth; every implementation decision is either anchored to it or explicitly flagged as our choice. The dominant failure mode is silent gap-filling — plausible defaults presented as the paper's content. That is neuroslop in code form, and this skill exists to make it structurally impossible.

Stage 0 — acquire the full text

Prefer arxiv-mcp tools (download_paper / read_paper) when available; otherwise WebFetch the abstract page and PDF.
Appendices, footnotes, table captions, and figure captions are first-class sources — missing hyperparameters usually live there, not in the method section.
If the authors published official code, locate it; it is a second, separately-tagged source — never a silent one.

Stage 1 — ambiguity audit, before any code

Classify every implementation-relevant decision; the tags below are this skill's specialization of the harness claim-typing discipline (epistemic boundary: confirmed ≠ associative):

Tag	Meaning	Obligation
`FACT §X.Y` / `Eq. N`	stated in the paper	anchor to the exact section/equation
`PARTIAL §X.Y`	mentioned but ambiguous	quote the passage verbatim, list the readings
`UNSPECIFIED`	paper is silent	our default + alternatives, never filled silently
`ASSUMPTION`	inference from context	reasoning stated next to it
`FROM-CODE`	taken from the authors' implementation	tagged as code-derived, not paper-derived
`QUESTION`	load-bearing gap that cannot be defaulted safely	ask the developer before coding — not a guess

A QUESTION blocks its code path until answered (mental model gate).

Stage 2 — implementation

Non-trivial decisions carry their §/Eq. anchor in the audit table; inline flags ([UNSPECIFIED], [ASSUMPTION]) appear only at load-bearing lines. These are constraint provenance the code cannot show by itself — not decoration, so they do not conflict with the no-decoration rule (invariant #9).
Variable names follow paper notation where that stays readable.
Standard components are imports: "standard transformer encoder" means a library call plus a dependency note, never a rewrite.
Scope is the core contribution only, unless training / data / evaluation pipelines are explicitly requested. No baselines, no infrastructure beyond what the contribution needs.

Read the full file on GitHub · 52 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 52 lines · 65 tokens per session scan A 446d2821ae7d

Subscribe to this mod's changes

paper-to-code is a skill published in the GitHub repository AlexShchuka/neuro-matrix (3 stars, last pushed 4mo ago), licensed MIT. It adds 65 tokens to every session and 857 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Claude Code Research Scientific ai-agents ai-code-review claude-skills harness harness-engineering

paper-to-code next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
paper-to-codeAlexShchuka/neuro-matrix	Skill	3	4mo ago	MIT	A	65	Claude Code
biopythonK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	65	Claude Code
neuropixels-analysisK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	99	Claude Code, Codex
onekgpdK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	143	Claude Code
pkpd-modelingK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	273	Claude Code
pydicomK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	52	Claude Code, Codex
dataladK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	147	Claude Code