WEBVTT

00:00:00.000 --> 00:00:03.345
Most teams adopted coding agents one engineer at a time.

00:00:03.795 --> 00:00:08.122
Nobody can say how much agent-written code reaches production unreviewed, whether

00:00:08.122 --> 00:00:11.649
the tests would catch it being wrong, or where the secrets end up.

00:00:12.099 --> 00:00:14.681
The work has two parts. First, an assessment:

00:00:14.681 --> 00:00:19.328
we read your pull requests, agent instruction files, tests and CI, read-only, and

00:00:19.328 --> 00:00:21.852
score each area with the evidence behind it.

00:00:22.303 --> 00:00:25.074
Then one day in your codebase with your engineers.

00:00:25.074 --> 00:00:29.563
In the morning, the team writes the agent instruction files and sets which files,

00:00:29.563 --> 00:00:31.891
secrets and commands the agents can reach.

00:00:32.341 --> 00:00:36.722
After lunch, an engineer does a real backlog ticket with an agent while the team

00:00:36.722 --> 00:00:37.160
watches.

00:00:37.610 --> 00:00:39.574
In the afternoon, checks go into CI:

00:00:39.574 --> 00:00:44.101
secret scanning on agent output, a PR size limit, and spot checks where the code is

00:00:44.101 --> 00:00:46.392
broken on purpose and the tests must fail.

00:00:46.842 --> 00:00:51.668
The day ends with a retro, an owner for the setup, and three numbers to measure for

00:00:51.668 --> 00:00:54.749
thirty days. The files and checks live in your repos.

00:00:55.199 --> 00:00:59.860
To start, email io@deemwar.com with your team size and the agents you use.
