Claude Code Skills for the late-start technical professional

The premise of Claude Code Skills is that you can package repeatable AI workflows as reusable, version-controlled commands. For a PhD or MS holder entering industry, this should feel familiar: you’ve spent years writing exactly these — SOPs, analysis pipelines, reproducible research workflows. Skills are the same idea, packaged for an LLM instead of a wet lab.

This collection brings together the open-source Skills most directly useful for late-start technical professionals doing real research work in industry. Every entry is a real installable Skill from a public GitHub repository — you can clone the source repo, copy the skill folder into ~/.claude/skills/, and start using it today.

How this collection is organized

The first cluster (literature, biomedical, clinical-trials, biorxiv, chembl) is the scientific search stack. All five live in yorkeccak/scientific-skills. They share the same Valyu semantic-search engine, just scoped to different sources. If you only install one, install biomedical-search — it’s the broadest. If you’re domain-specific, the specialized ones return tighter results.

The second cluster (anti-slop, writing-clearly, content-research-writer, writing-plans) is the writing and quality stack. These are the editing layer that turns AI-drafted text into something you’d put your name on. Anti-slop is the most-installed skill in the public registry (65k+ installs) for a reason: nobody wants visibly AI-generated work attached to their reputation. Writing-clearly is the constructive companion. Together they cover most of what you need to ship AI-assisted writing without losing your voice.

How to use them together

A working pattern for industry research scientists:

1. Scope a research question with /biomedical-search or /literature-search
2. Draft a synthesis using your normal Claude Code conversation, leaning on the search results
3. Run /anti-slop and /writing-clearly-and-concisely on the draft before showing it to anyone
4. Use /writing-plans if the work needs a structured plan before code

That’s a four-skill workflow that compresses what used to be a half-day exercise into about ninety minutes — and the output is yours, not a generic AI summary you’d be embarrassed to ship.

What this collection is not

These are starting points. Every Skill in this list will need 20–30 minutes of system-prompt customization to fit your specific domain. A genomics researcher and a quant will calibrate /biomedical-search differently. The defaults are usable; the custom versions are valuable.

What’s missing (and on the roadmap)

• A Skill for converting an academic CV section into industry-formatted resume bullets — currently doable manually with Claude Code but not packaged as an installable Skill
• A regulatory-affairs companion to /clinical-trials-search that pulls FDA guidance documents
• A statistical-methodology checker for late-start scientists writing analysis plans

If you’ve found or built something that fits these slots, the newsletter is the right place to share it.