Working with agents
Your agents will work the way your team does.
An agent’s output is only as consistent as what it’s given to read. I’ll write that down with you: the words your team uses, the conventions it keeps, and what “ready” means. I run this site the same way: How this site is built quotes the glossary, the decision records and the labels that route work between me and an agent.
What you get
- Specs an agent can run, and a triage flow that says what’s ready for an agent and what needs a human
- Skills, context files and decision records that carry your conventions into every session
- Automation on cheap, always-on hardware, so routine work is done by the time your team logs on
Working together
In an Embedded build, I join your team for the long haul. I’ll start with how you use AI today, even if that’s hardly at all. Then I’ll put the specs, skills and automation to work on your backlog, and measure how much of your real work goes through agents, so you can show what the spend is buying.
Evidence
This site, built with agents
Agents wrote nearly all of this site’s code, and I reviewed it. They built it from issues I’d triaged with them and decisions I’d written down first, and I ruled on whatever they couldn’t decide.
109 pull requests in the build’s first week
Read the write-upSay hello
Email hello@phil.dev. I read everything that isn’t a sales pitch.