Rendered from docs/decisions/0032-q32-which-test-the-self-audit-label-states.md in the Headwater corpus. Every document on this half of the site is typed by the taxonomy the descriptor names: corpus.json.

Q32 — Which test the self-audit label states

Context

The build order of this repository states one value rule, in .claude/commands/next-run.md. Before any work starts, somebody names the reader who is not this repository. Where the only party better off is this corpus, the work is not eligible for an iteration.

Two labels on the issue tracker carry that rule. adopter-blocking marks work an outside adopter cannot proceed without. self-audit marks the other side of it, and the label sorts work out of every selection rule the build order reads.

The description of self-audit named two tests at once. It read as work found by running the engine over this repository with no reader outside it. The first clause is a test of provenance, which is how the finding arrived. The second is a test of the reader, which is who is better off when the work is done. The two partition the board differently, and nothing said which one decides.

The conflation costs work in both directions. A defect found by running the engine over this corpus can block an adopter, and the provenance test hides it. A defect found by reading an issue from outside can serve nobody but this corpus, and the provenance test admits it.

Decision

self-audit states the reader test. The label description reads No reader outside this repository. How it was found is not the test.

The reader test is the one the value rule already states. The rule asks for the reader who is not this repository, and it asks for nothing about the instrument that produced the finding. A label that carried the provenance test would sort the board by a fact the rule does not read.

Provenance is a poor test on its own terms. Almost every finding in this repository arrives from running the engine over this repository, because that is the only corpus there is. A test that nearly every candidate passes separates nothing.

The label marks a wait and never a fault. self-audit says that the work has no reader outside this repository today, so it stands open and something else runs first. It says nothing about whether the work is right.

Consequences

A finding found by the engine over this corpus is not self-audit when an adopter needs it. The label comes off, and the work competes on the same footing as anything else. adopter-blocking is the label that then applies, and it sorts above everything in the selection rules.

A finding found anywhere else is self-audit when no outside reader exists. The origin of the report does not exempt it.

The two labels stay the product owner's. .claude/agents/headwater-product-owner.md writes both, and no other agent applies either one. This record settles which test one of them states, and it settles nothing about who applies it.

The value rule keeps its single statement. .claude/commands/next-run.md holds the canonical wording, and every other file cites it. This record adds no second copy of the rule and reads the label description against it.