Separation of duties

What it is

Separation of duties means a task's review or merge stage requires a different actor than whoever handled an earlier stage. The person who wrote the code isn't the one who approves it; the agent that coded a change isn't the one that verifies it. It's a structural check. Trusting any one actor more or less has nothing to do with it, and it applies the same way whether both actors are human, both are agents, or one is each.

Why it matters to you

A single actor checking its own work, whether that's a person or an agent, is the weakest possible review: an agent that misunderstood the task in a given way will make the same mistake reviewing it too. Requiring a second, independent set of eyes is what actually catches a wrong answer before it ships, and it's the kind of governance a solo desktop tool was never built around, because there was never a second party in the loop to require it of.

How it works in CodeHerder

CodeHerder can require a role or a named person distinct from whoever advanced a task's prior stage, enforced at the workflow level rather than left as an honor system, so it can't be quietly skipped when a team is moving fast. This sits alongside CodeHerder's other shipped verification, from per-stage required fields to checklist and dependency gates, all laid out on the verification page.

• Verification →

What people get wrong

It's tempting to think separation of duties only matters at large companies with compliance requirements. It matters just as much for a two-person team: it's the difference between a change that got a second opinion and one that only ever had its author's own judgment behind it. The value doesn't come from the size of the team, it comes from a second, genuinely independent look existing at all, which a solo team can still get by having an agent review another agent's work.