# Success criteria

> Success criteria are the specific, measurable conditions an agent's work must meet to count as done, written before the work starts so the result can be checked instead of judged by feel.

Each criterion is one observable outcome that can fail. The test: at the deadline, could two reasonable people disagree about whether it was met? If yes, it is still an activity. "Review the runbook" fails; "every runbook step has a named owner by 5 September" passes. Spec-driven tools such as Kiro and GitHub Spec Kit call the same idea acceptance criteria, and evals are how the work gets checked against them.

**How I use it:** I made success criteria a required section of the skill that spawns my agents, so every new agent starts from a template that asks for two to four of them before the agent exists.

**Source:** Anthropic's guide to [defining success criteria and building evaluations](https://platform.claude.com/docs/en/test-and-evaluate/develop-tests); in spec-driven development the same idea is [acceptance criteria](https://www.atlassian.com/software/jira/guides/agentic-engineering/spec-driven-development-jira).

Used in [Spawn every agent with a spec, a plan, and the same workspace](/garden/agent-spec-plan-workspace).