Rendered from docs/probes/a-session-names-the-status-a-settled-decision-carries-in-its-pull-request.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 names the status a settled decision carries in its pull request

Task

You add a new decision record to this repository. You consider the decision settled, and you open a pull request for it. A person reads the pull request before it merges.

Which value does the status field in the front matter of that document carry in the pull request, draft or current? Which decision record of this repository rules it? Name that record by the identifier in its id field.

Answer on one line with the value, one space, and the identifier, and with nothing else. Your whole final message is that line: no sentence around it, no justification before it, and no Markdown emphasis on it.

Expectation

The terminal answer is current HW-DR-0052, the one expected value of the closed set below.

answers: [current HW-DR-0052, draft HW-DR-0052]
expected: [current HW-DR-0052]

The closed set holds both values that the task offers, each with the identifier of the ruling. So a recorder records draft HW-DR-0052 as that line and not as no answer, and the grade finds it wrong. A line that names another identifier, or no identifier, is outside the set. The recorder records it as no answer, and the grade finds it not satisfied. headwater probe plan refuses a probe whose expected values are every value of its set, because every answer would then pass (#1229).

The convention is HW-DR-0052. An author writes the state that the document will hold once the branch lands, and the merge activates it. headwater new writes draft, because the lifecycle regime opens at that state.

Why the task asks for the identifier. The first task of this probe asked for the value alone, and the model answered it from what it already knew. The campaign of 2026-09-28 recorded 30 sessions in each of three arms. Those transcripts are now deprecated, because this change moved the read set of their probes, so the figures below are the reason for the change and not a current measurement. The present arm of the campaign tier answered current in 21 of 30. The absent arm of that tier answered current in 30 of 30, and the absent arm of the documentation tier in 28 of 30. None of the 90 sessions made a tool call. So a session that read nothing still gave the expected value, and the probe did not measure the documents. The identifier of the ruling is a fact that only this repository holds. In the campaign present arm, a session gets it without a read, because the always-loaded description of the headwater-authoring skill names HW-DR-0052 and status: current. On 2026-09-30, 30 of 30 present-arm sessions answered with no tool call. The absent arms have no .claude/, so a session there gets the identifier only when it reads the ruling or a file that names it.

Why the probe does not grade opened on the ruling. An opened expectation names the ruling through examines, and the documentation tier removes docs/. So headwater probe plan --tier documentation refuses a probe whose examines target is under an ablation entry. That grade would keep the probe in the campaign tier alone and lose the comparison between the two absent arms that HW-OBL-0223 needs. An identifier in the closed set keeps the probe answered, and both tiers admit it.

This probe is for the documentation tier. Its answer is stated under docs/, and the campaign tier's absent arm keeps docs/. So the difference between the two absent arms is the part of the rate that the documents cause.

Where else the answer is written, and which arm holds it. This document states the answer, so it is part of the instrument that .headwater/probe.yml names, and every arm of every tier removes it. The name of this document states no answer, because engine/crates/census/fixtures/corpus.census lists it and that file sits outside docs/. The same census fixture lists the file name of the decision, and that name states the ruling. So the documentation ablation removes the census fixture with docs/. The hand-built tutorial page site/tutorial/index.html states the ruling in one step, and the same ablation removes it. The authoring skill under .claude/ also states the ruling, and both absent arms remove it.

In the present arm, this probe measures the leak string in the description of the authoring skill (#1472). The description of that skill states the status and the identifier of the ruling, and a harness loads the description into every session. So a present-arm session can answer before it reads a document. In the #1384 re-run, 30 of 30 present-arm sessions of this probe answered with no tool call. The leak string stays on purpose, because the description is how the skill delivers the ruling, and removing it would change the treatment that the present arm measures. .headwater/probe.yml declares the leak strings of this probe under leaks: and keeps them under leaks_kept:. sh tools/probe/seal.sh --leak prints the leak string as kept. A campaign runs this probe on a line of its own and never pools it into the rate of the sufficiency category. Only the owner can choose to remove the leak string instead.

A hint stays in both absent arms, so the difference is a lower bound. A test comment in engine/crates/scaffold/tests/pipeline.rs calls a move from draft to current the author's promotion by hand, as HW-DR-0052 has it. That is not the ruling, but a session that reads it can reach the whole expected line without a document. The identifier also appears in a list of claimed identifiers in tools/repo/id-store-fixtures.sh, with nothing that ties it to the status. Both files sit in both absent arms, so they lower the difference between the arms and cannot inflate it.

An answered probe names no document through examines, so the documentation tier admits it.

HW-OBL-0223 records the unmeasured claim. This probe supplies an instrument for it and does not supply an observed result.