A test pins the run-cap park: it is noted and asked once per window, and a quoted reason is cut at a sentence #6904

src/crew/guard.ts decides when an issue has used its runs (capReached) and parks it once (capNoted). quoteCut trims what a park quotes. None of them is named in a test.

Done:

• [ ] test/crew-guard-cap.test.ts: an issue under its cap is not capped; at its cap it is. A check-failure run counted elsewhere doesn't count here, as the docblock says.
• [ ] After the cap note, a review or a reader's comment does not start a second note. A director's word starts a new window.
• [ ] quoteCut returns short text unchanged and cuts long text at a sentence end inside the window when there is one. It never exceeds its max, and never splits a surrogate pair.

This is a new test file only. Change code only where the test finds a hole, and say so in the pull. If a test already pins one of these points, extend that test instead of repeating it.

Files: test/crew-guard-cap.test.ts

Before pushing: bun x tsc --noEmit and the test you add. Tests pass now in and never read the wall clock.