Intent debt

Intent debt is the gap between the goals and constraints that should be steering a system and what actually got captured in writing, so the artifacts stop reflecting what you meant to build.

It is one of three layers in Margaret-Anne Storey's model of software health, alongside technical debt (which lives in the code) and cognitive debt (which lives in a team's shared understanding). Intent debt lives in the artifacts. In an agent spec it is the part of an agent's real purpose that never made it onto the page. My rule is to close what a short conversation can and log the rest as an explicit Open question, with the assumption I made and who can settle it, rather than leaving a quiet guess in its place. A spec that says "we do not yet know what good looks like here, assuming X for now" carries its intent debt in the open, which is more honest than a made-up metric, because debt you can see is debt someone can pay down.

How I use it: I name it in an agent spec instead of hiding a guess, so the missing purpose stays visible and payable. As Addy Osmani puts it, an agent cannot generate intent, so it is the one input that has to come from me.

Source: intent debt is one layer of Margaret-Anne Storey's Triple Debt Model (technical, cognitive, and intent debt), discussed by Martin Fowler and developed for AI agents by Addy Osmani.

Used in Spawn every agent with a spec, a plan, and the same workspace.