Your tests run, but nothing measures which lines they reach

NO_COVERAGE_TOOL · exit 10 · an agent can clear this

The suite runs and a receipt can prove it passed. What it cannot prove is that it executed the lines you changed, which is the question this policy asks. That needs a coverage tool alongside the runner.

Clears when

A coverage tool is installed and the next receipt carries a report.

Commands

<install the coverage tool for your stack, e.g. npm i -D c8>
relayevals next --format json

Every state, with its exit code, is listed in the documentation. An agent that reaches an unfamiliar state should run relayevals next --format json and follow the single action it returns.