
claude-skills
Mike Frank's Claude Code skills (checked out as ~/.claude/skills)
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up claude-skills (git-clone project) into my current project. Found on https://claudeers.com/claude-skills-12 Repo: https://github.com/mcfrank/claude-skills Homepage/docs: — Detected install method: git-clone → git clone https://github.com/mcfrank/claude-skills Category: education. Platforms: cli. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: unknown; community-verified: false. Confirm the source before running anything.
git clone https://github.com/mcfrank/claude-skills
// compatibility
| Platforms | cli |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | MIT |
| Pricing | open-source |
| Language | Python |
claude-skills
Mike Frank's Claude Code skills. This repo is checked out directly as
~/.claude/skills, so every skill here is available in every local Claude Code
session. Frontmatter description fields are matched against your request to
decide when to auto-load them.
Using these in a cloud session
Add this to the cloud environment's setup script:
git clone --depth 1 https://github.com/mcfrank/claude-skills ~/.claude/skills
Or copy a single skill into a repo's .claude/skills/<name>/ to scope it to that
project.
What's here
Stan / Bayesian model fitting
gcp-large-stan: fit big Stan models on GCP n2d-standard-* VMs (sizing, SMT-aware threading, compile-flag pitfalls, disk, detached launches).stan-mixing-cold-start: diagnose and recover from non-convergent Stan fits; in family-build workflows the usual cause is warm-starting from a degenerate source posterior.psis-loo-chunked: PSIS-LOO on very large log_lik matrices without segfaults (chunking is exact).
Remote compute
remote-vm-gotchas: SSH and long-job footguns (scp into a running script, pkill self-matching, detached launches,/tmpfilling up).
Research workflow and evaluation
adversarial-protocol-review: "what could manufacture this result?" review of experiments, evals, and analyses.pre-submission-review: required AI review of a lab paper before submission; read-only on the manuscript, with a findings file for the author to decide on.reproducibility-audit: computational reproducibility audit of a published paper (OSF fetch, prereg check, Quarto report template, error checklist, robustness toolkit).rmd-qmd-manuscript-collab: collaborating on human-written manuscripts (code, YAML, templates, bib) without touching prose.quarto-book: building and publishing a Quarto book of analysis notebooks.
Data infrastructure and R packages
redivis: Redivis datasets, uploads, tokens, clients, releases.shiny-to-datapage: migrating a Shiny/Django data site to Quarto + Observable over Redivis.cran-safe-release: keeping internet-dependent R packages on CRAN.
Lab projects and datasets
wordbank: the Wordbank project, its new Redivis/Quarto stack, and wordbankr.cdi-vocabulary-data: CDI measurement, Wordbank access, IRT, vocabulary growth.babyview-dataset: the BabyView head-camera corpus, its pipeline, releases, and storage tiers.
Local-only skills
A few skills hold internal or laptop-only details (program admin, course
internals, cluster auth), so .gitignore keeps them out of this repo. Published
skills can likewise keep sensitive specifics in a gitignored private.md next
to their SKILL.md.
Project-local overrides
A project's .claude/skills/<name>/SKILL.md takes precedence over a global
~/.claude/skills/<name>/SKILL.md of the same name. Use project-local for
specifics (VM names, repo paths, variant lists) and global for transferable
lessons.
Adding new skills
Create ~/.claude/skills/<name>/SKILL.md (a bare .md at the root is NOT
loaded) with YAML frontmatter:
---
name: my-skill-name
description: One sentence saying what it covers and when to use it.
---
# Body content in Markdown
The description is the auto-match trigger. Write it as the answer to "when
would I use this?". Then commit and push; if the skill holds anything that
shouldn't be public, add it to .gitignore first.
// faq
What is claude-skills?
Mike Frank's Claude Code skills (checked out as ~/.claude/skills). It is open-source on GitHub.
Is claude-skills free to use?
claude-skills is open-source under the MIT license, so it is free to use.
What category does claude-skills belong to?
claude-skills is listed under education in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/claude-skills-12)
// retro hit counter
[](https://claudeers.com/claude-skills-12)
// reviews
// guestbook
// related in Education & Learning
Skills for Real Engineers. Straight from my .claude directory.
Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.
Learn it. Build it. Ship it for others.
DeepTutor: Agent-native Personalized Tutoring. https://deeptutor.info/.