Rendered from docs/probes/a-session-answers-from-the-register-when-the-task-names-the-list-it-replaced.md in the Headwater
corpus. Every document on this half of the site is typed by the taxonomy
the descriptor names: corpus.json.
A session answers from the register when the task names the list it replaced
Task
This project keeps its open questions, closed and redirected. Say what it decided on the open question of where scent lives, and name the record that decided it.
Expectation
not_opened over HW-REG-open-questions. The session answered and never opened the superseded list of open questions.
This probe stands beside the register probe and does not replace it (#1718). The campaign of 2026-10-03 recorded that probe at 30 of 30 in the present arm and at 30 of 30 in the absent arm. Its task asks where scent lives, and the file name of the decision, docs/decisions/0020-where-scent-lives.md, holds those words. So a name search in either arm reaches the decision first, and no session had a reason to open the superseded list.
This task uses the title of the superseded list. docs/spec/09-open-questions.md carries the title Open questions, closed and redirected, and its file name holds open-questions. Both phrases are in the task. So a name search for the words of the task reaches the superseded list as early as the decision. The question is still Q20, so that the wording of the task is the one thing that differs from the original.
This probe grades avoidance and never correctness, as the original does. Both registers carry the Q20 line, so a right answer is evidence of nothing about the route. A session that opens the superseded list, follows its link, and answers correctly is a miss here.
Why it should separate the arms. In the present arm, a hook runs headwater route on the prompt and prints its pointers before the first tool call. Over this task, measured on 2026-10-10 at the commit this probe lands on, those pointers name the decision register docs/spec/09-decisions.md and not the superseded list, once the probe shelf is removed. The orientation skill also sends a session to headwater explain, which reports status: superseded for the list. In the absent arm, a session has only a search, and the words of the task name the superseded list.
What this probe expects, stated before any paid run. In the present arm, 24 of 30. In the absent arm, 16 of 30. These are predictions and not measurements, and a campaign decides them. If both arms reach 28 of 30 or more, the title of the list does not draw a session to it, and this probe is at the ceiling too. If the absent arm is the higher of the two, the layer sends a session into the list rather than past it, and that is a result too.
A session that did nothing is not a session that avoided anything. The grader refuses not_opened where the transcript records no tool call at all. So a refusal leaves the denominator and never the numerator.
HW-OBL-0197 is discharged. This probe adds a harder reading of the claim it recorded, and it changes nothing about that record.