Skill(.md)s, context engineering, and a way to take the garbage out
by Robert Tomb | at Minnebar20 | 12:30 – 1:10 in Harriet | View Schedule
SKILL.md files are a great way to export knowledge from one engineer to another. Think of them as training docs that someone can read, but if they don't want to, they can assign their assistant, or copilot, to read them and do the work.
Sharing knowledge (context) across larger engineering organizations that are engaged in agentic engineering (go ahead, call it vibe coding, but then we can't be friends), should make everyone more productive, until it doesn't. Done right, big win. Done wrong, failures may not be recognizable as such until well after you shipped what you thought was working code.
In this session, I'll cover:
- general benefits of SKILL.md files,
- how my teams are using SKILL.md files at Shipt to share context with one another
- Cite an academic study measuring the effects of bad context on coding assistant outcomes
- Tell you why I believe you should audit your context if you are not sure if your SKILL.md files might be conflicting
- Show a SKILL.md that checks other skill files (yo dawg, I heard you like SKILL.md files)
- Tell you how you can do the same
Plead with you to stop thinking of SKILL.md files a "just markdown"
Robert Tomb
Well, a simple tagline has been: "a nerd in the twin cities who has a family, a bicycle (or so), and some other stuff," which is pretty accurate.
Beyond that, I currently work in Minneapolis as a Director of Engineering for Shipt, by way of a couple of years in Target's tech organization. Prior, I'd worked in online advertising where I got my start in the SaaS world. Even before that, it was all on-prem software in the ERP world. I've helped build remote, local, and hybrid teams starting as far back as 2006. I like to build on those experiences to help new engineering managers grow their teams.
You may recognize me as a volunteer from previous Minnestar events, and now I'm helping out on the board. Volunteering with Minnestar might be the longest-running volunteer stint in my life.
Links:
Are you interested in this session?
This will add your name to the list of interested participants. It will help us gauge interest for scheduling purposes.
Interested Participants
- Casey Helbling
- Jamie Thingelstad
- Chris Warren
- Jachin Rupe
- Zack Steven
- Robert Tomb
- Sam Grumdahl
- Mark Norgren
- Dan Wick
- Benjamin Ortega
- David DeCesare
- Scott Vlaminck
- Matt Decuir
- Sarah Olson
- Matthew Chimento
- James Molohon
- Joseph Heil, Jr
- Matt Eskolin
- Alex Wohlhueter
- Peter Clark
- Sean Casserly
- Mark Mankey
- Dan Letsche
- Shalanah Dawson
- Jam Leomi
- DeLonn Crosby
- Kelly Heikkila
- Andrew Comfort
- Nate Anderson
- Bruce Adelsman
- James Hunt
- Brett Saxon
- Misha Hawthorn
- Ben Hymans
- Thibaud Cholat
- Eric Sirkin
- Paul Esch-Laurent
- Irene O
- Joe Lencioni
- Alex Frescoln
- havi
- Matt Glatzel
- Jeff Peterson
- Matt Rushin
- Kendrick Holden
- Bo Driscoll
- Chris Barber
- Cary Christopherson
- Mitch Bliven
- Nathan Perfetti
- Benjamin Schatz
- Josh Benner
- Walter Ethan Lick Eagle
- Aniruddha Ghosh
- Angela Peterson
- Nate Yourchuck
- Tyler Perkins
- John Evans
- Steven Bertsch
- Andy Ganoe
- Justin Kaster
- Jesse O'Neill-Oine
- James Greene
- Mark Mykkanen
- Koy Payne
- Jesse Anderson
- Tina Trinh
- Tim Brunelle
- Daniel Hokanson
- Chris Rothwell
- Matt Brinster
Similar Sessions
Does this session sound interesting? You may also like these:
-
Building a Knowledge Graph Your AI Can Actually Use
-
AI Told You So Again: How I Built a $1B Unfair Advantage ... And So Can You (Meet OpenClaw)
-
10x'ing Myself and My Team: Leading With Agents, Not Just Using Them
by Andy Ganoe -
What the fuck are passkeys and why are they everywhere now?
by Dan Lew -
Train AI To Be Your Employee