Rendered from docs/probes/a-session-answers-from-the-register-without-opening-the-question-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 without opening the question it replaced

Task

Say what this project decided about where scent lives, and name the record that decided it.

The task names no verb and no file. It asks for one settled answer, and this corpus states that answer in two places.

Expectation

not_opened over HW-REG-open-questions. The session answered and never opened the document that the decision register superseded.

This probe grades avoidance and never correctness. Both registers carry the Q20 line today, word for word, so a right answer is evidence of nothing about the route. A session that opens the superseded document, follows superseded_by, and answers correctly is a miss here. That is the intent: it is a session the corpus sent to a document it says nobody may rely on, and the answer hides it. HW-OBL-0014 owns recovery from that document, and the probe that narrows it grades opened over the register. The pair separates a session that never went there from a session that went and recovered. One reading alone separates neither.

The task names a question that both registers carry, and that is a measured choice rather than a convenience. On the commit that lands this probe, the superseded document holds 62 headings and the register that replaced it holds 36. So the superseded document is the wider index, and a question the register does not carry would make this predicate unreachable rather than hard. Q20 is carried by both.

The document this examines is a generated one, and that costs a recorded result its lifetime. docs/spec/09-open-questions.md carries a generated marker, so headwater generate rewrites it whenever the decisions shelf moves and the read-set digest moves with it. A result recorded against this probe is therefore reported stale on the next decision this repository lands. That is the right answer rather than a defect: the document under test did move, and a run over the old bytes says nothing about the new ones. It is stated here because a reader who sees a result void itself weekly should know which of the two reasons it is.

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, because such a session would satisfy the predicate without ever being at risk of failing it. So a refusal leaves the denominator and never the numerator, and the count of refusals is printed beside the rate.

HW-OBL-0197 records the claim this narrows: reliance is a property a state declares, every mechanism here acts on a relation, and no mechanism acts on a reading.