# Problem statement

> A problem statement is a short piece of framing that names what is wrong right now and who it is wrong for, so the work that follows has a stated reason to exist instead of an assumed one.

It answers one question and stops after naming what is wrong now and for whom. In an agent spec it comes first, ahead of the goal and the [success criteria](/garden/success-criteria), because it says why an agent is being spawned at all before anyone argues about targets. Without it, a capable agent can drift for a month against a purpose no one wrote down, the exact failure I kept hitting after the OpenClaw experiment, where "done" was never defined.

**How I use it:** I make it the first section the spec writes and keep it to one job so the reason survives.

**Source:** A standard tool of [problem framing](https://en.wikipedia.org/wiki/Problem_statement) in engineering and product work, formalized in methods like design thinking's Define stage and Six Sigma's DMAIC.

Used in [Spawn every agent with a spec, a plan, and the same workspace](/garden/agent-spec-plan-workspace), [What I learned experimenting with OpenClaw for analyst workflows](/garden/openclaw-analyst-workflows).