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.
npx skills add queelius/claude-anvil --skill notebook-paired-with-prosegit clone --depth 1 https://github.com/queelius/claude-anvilWrote 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.
[](https://agentmods.dev/skills/queelius/claude-anvil/notebook-paired-with-prose)<a href="https://agentmods.dev/skills/queelius/claude-anvil/notebook-paired-with-prose"><img src="https://agentmods.dev/badge/skills/queelius/claude-anvil/notebook-paired-with-prose/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.
<a href="https://agentmods.dev/skills/queelius/claude-anvil/notebook-paired-with-prose"><img src="https://agentmods.dev/badge/skills/queelius/claude-anvil/notebook-paired-with-prose.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00058 | $0.00894 |
| Opus 5 | $0.00029 | $0.00447 |
| Sonnet 5 | $0.00012 | $0.00179 |
| Haiku 4.5 | $0.00006 | $0.00089 |
Grade A, and why
notebook-paired-with-prose 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 11d 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.
How it starts
The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Notebook Paired with Prose (bookwright)
In a technical textbook, paired notebooks are not optional decoration. They are empirical verifiers: a chapter that claims a formula or worked example has a notebook that COMPUTES the formula and shows the empirical result matches. Broken notebook means broken chapter.
When notebooks are required vs optional
Required:
- Any chapter with a worked numerical example. The notebook must compute the same example and match.
- Any chapter with an algorithm. The notebook implements it and runs it.
- Any chapter with an empirical claim ("the SE is roughly 0.05"). The notebook simulates and verifies.
Optional or omitted:
- Theory-only chapters (no worked examples, no algorithms). Bernoulli's chapters 4, 13, 17 had no notebooks; the master spec marks them theory-only.
- Bridge chapters that exist only to connect Parts.
The master spec's chapter outline marks which chapters have notebooks. The per-chapter plan's notebook task spec lists the content + numerical-sanity targets.
Numerical-sanity-target convention
The plan lists explicit targets like:
- "Empirical FPR at $\fprate = 0.01$, k=7: within plus-or-minus 0.005 of theoretical 0.0078."
- "Wald CI coverage at $n_0 = 200, \fprate = 0.05$: in $[0.93, 0.97]$ band."
The notebook-author agent computes these and compares. Match = pass. Miss = flag for user review (could be the prose is wrong, could be the notebook is wrong, could be sampling noise).
Exec-from-fresh-kernel requirement
Notebooks MUST execute end-to-end from a fresh kernel with no errors. The notebook-author runs:
- Python+uv+Jupyter:
cd <project>/notebooks && uv run jupyter nbconvert --to notebook --execute <name>.ipynb --output <name>.ipynb - R+renv+RMarkdown:
cd <project>/rmd && Rscript -e 'rmarkdown::render("<name>.Rmd")' - Quarto:
cd <project>/qmd && quarto render <name>.qmd
Exit code 0 is required for commit.
Common gotchas
- Python: forgetting to seed the RNG produces non-reproducible outputs. Always
np.random.default_rng(seed=42)or similar. - Python: dependencies must be in
pyproject.toml. New dependencies requirecd <project> && uv add <pkg>before re-executing. - R:
renv::status()should be clean; new packages requirerenv::install("pkg")thenrenv::snapshot(). - Quarto: cell-by-cell execution depends on engine; if mixing R + Python, configure the engine carefully.
- All stacks: explicitly seed any randomness. Markdown cells should describe what the code cell will do, not just narrate it.
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.
- 11d ago First seen · 70 lines · 58 tokens per session scan A 4d29d0ed2727
notebook-paired-with-prose is a skill published in the GitHub repository queelius/claude-anvil (2 stars, last pushed 2mo ago), licensed MIT. It adds 58 tokens to every session and 894 once invoked, about $0.0003 per session on Opus 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.
Other skills, from other repositories
lean-check
Formalize a self-authored lemma or theorem in Lean 4/mathlib and require a clean lake build without sorry. Use when the mathematical claim can be stated faithfully and machine-checked. For numerical falsification or symbolic algebra, use $numerical-check or $symbolic-check.
beamer-academic
Generate high-quality academic Beamer slides from a thesis/paper (PDF, Word, or LaTeX source). Compatible with Claude Code and Codex. Supports thesis defense, proposal presentations, and conference talks. Built-in layout library with 13 professional page types, 5 color schemes, and interactive editing loop. Use when…
math-proof
Write clear, detailed mathematical proofs for academic papers. Use when the user asks to prove a result, derive an equation, justify a claim analytically, or expand a proof sketch into a full proof. Also trigger on "prove", "show analytically", "derive", "justify mathematically", or "write a proof".
jupyter-notebook
Create, scaffold, or refactor Jupyter notebooks (.ipynb) for experiments and tutorials. Prefer the bundled templates and the helper script (newnotebook.py, also exposed as newnotebook) to generate a clean starting notebook instead of authoring raw notebook JSON.
explain
Use when user invokes /explain with a file path, directory path, function/class name, or natural language concept. Also triggers on "explain this", "how does X work", "walk me through". Produces a structured, layered explanation of what the code does, how it connects, and where to start if you need to change it. Do…
reinforcement-learning
Reinforcement Learning best practices for Python using modern libraries (Stable-Baselines3, RLlib, Gymnasium). Use when: Implementing RL algorithms (PPO, SAC, DQN, TD3, A2C) Creating custom Gymnasium environments Training, debugging, or evaluating RL agents Setting up hyperparameter tuning for RL Deploying RL models…