Skills for Real Engineers

Skills for Real Engineers is Matt Pocock's published agent-skills directory: small, composable instruction files that install into Claude Code, Codex or another agent, written to be adapted rather than adopted whole. Most of the collection serves software work, but several skills are education instruments that were built for teaching and questioning rather than for code.

The clearest example is teach, which runs over multiple sessions and treats the working directory as a stateful teaching workspace, so a learner's progress and open questions persist between conversations instead of restarting each time. grill-me and its underlying grilling primitive interview the user relentlessly about a plan until every branch is resolved, which is Socratic questioning as a reusable procedure rather than a conversational mood. wait-what responds the moment a message fails to land by re-pitching it in plain language with the missing context, a move any instructor recognizes from watching an explanation miss the first time. writing-for-agents covers how to write documents that an agent can follow, which is now the practical form of writing instructions for AI course material. to-questionnaire converts a decision into a questionnaire for whoever can answer it, handoff compacts a conversation into a document another agent can continue from, and wizard generates an interactive walkthrough for steps only a person can perform.

What to know before adopting it

Everything is open source under MIT and free to take as a starting point for local AI literacy or self-directed study. The adoption is large and fast-moving: roughly 269,000 stars and 22,000 forks when checked in September 2026, with the last commit that day. The caveat is scope. The skills assume an engineer's working context, they are organized into engineering, productivity, deprecated and in-progress buckets, so some entries are explicitly unfinished or retired, and the README doubles as a newsletter sign-up. Treat the teaching and questioning skills as the transferable part, and expect to rewrite any of them before putting one in front of students.

Connected Concepts

Socratic Method, Self-Directed Learning, Metacognition, Prompt Engineering, Generative AI, Pedagogical Agent, AI Literacy, Human AI Collaboration, Open Source, Teacher AI Competency