Name the whole default
Record the model, agent, instructions, skills, tools, permissions and checks as one versioned setup, like backend-tools@18. When something regresses six months later, you can see which version produced which results.
For Platform & DevEx
What should Platform teams standardize for coding agents?
You already version CI config, toolchains and dev environments. The coding agent setup every team shares changes just as often, usually without a version number or a test. Fullbeam gives each version a name and a benchmark.
In practice
What this looks like on your code
Every result links back to the setup, the tasks and the commit it came from. If something wasn't measured, you see that too, instead of a number that looks complete.
Benchmark before teams adopt it
Run the current default and the new one on tasks from the repos they'll affect. Same starting commit, same tests, same rules for what counts as a pass.
Keep more than one good default
Routine maintenance and database migrations may need different setups. Your standard can include several, each with a clear list of the work it covers, instead of one choice forced on every repo and team.
Turn failures into tasks
When an agent change goes wrong and someone merges the fix, that PR can become a task. The next default has to pass it before another team gets it.
Got a coding agent change coming up?
Fullbeam tells you which kinds of work should move to the new version. GitOps, MDM or your internal platform still deploys it.
Your current setupThe change you want to tryPRs you already mergedAn answer for each kind of work