Quill Builder Bar — Impl Gate

This page is the public surface of Quill’s impl-gate Builder bar. It is consulted one criterion at a time: look up the rule or the catalog entry in front of you, and leave with what fires it, what must not, what a report of it must cite, and whether it stops the gate.

What this bar is

This is the Builder actor’s bar at the impl gate — the gate that grades a finished document against its frozen .feature.

It runs once per document, not once per scenario. The per-scenario checks run against the frozen .feature; this bar runs once over the whole document.
It unions onto SDD’s generic conformance bar rather than replacing it. The generic requirements — checks derived from the frozen .feature, no green-by-tampering — still apply in full.
Every criterion in it is a relation between two passages. Each occurrence is well-formed when read on its own, and only the pair fails, which is why a per-scenario check cannot carry any of them.

What the bar carries

The bar carries two instruments: one enumeration rule, and a judged catalog of nine entries.

Why the two instruments split the way they do is argued at the doc-eval model. This page carries what each instrument is and how it is run.

The enumeration rule

A route must reach every option the document itself named. Where the document enumerates a set — its mechanisms, its arrangements, its kinds — and a later passage routes a case across that set, an option absent from the routing fires this rule.

An option absent from the routing is a defect, not a simplification. The dropped member is usually the one introduced between the two the routing keeps, because the routing was drafted from an earlier version of the passage, before the set was complete.

Citation. A report of this rule must quote the passage that enumerates the set and the routing that skips a member. Both, with their locations.

Verdict: blocker. A failure of the enumeration rule blocks the gate. It is not an advisory finding — this is inspection, settled by comparing two enumerable things, so it does not wait on calibration the way every catalog entry does.

The defect catalog

The catalog names prose defects. It has nine entries in three groups, and the groups are named by what the defect does to a reader:

Group	What the defect does to a reader
A	the reader cannot retrieve what the passage assumes
B	the passage misrepresents what the reader already has
C	the document disagrees with its own spec

Each entry states the case that must not fire it — its near-miss. An entry without a near-miss is a style opinion with a rubric attached.

It detects defects; it never certifies quality. Good prose is unbounded and cannot be enumerated; bad prose recurs in a small number of nameable shapes. A document with zero findings is not thereby endorsed — a clean judged pass says no listed defect was found, and nothing more.

Group A — what the reader cannot retrieve

Throughout this group, the reader means a reader on one declared control-flow path. Content elsewhere in the document licenses nothing if the path does not reach it.

Citation this group owes. These findings are negatives — they claim something is absent — so the absence must be demonstrated over a named path: quote the passage, name the path, and list what that path traverses before it. “I did not find it” is not a finding.

A1 — Unresolvable presupposition

Fires on: a passage treating X as given — presupposing it rather than asserting it — when nothing the reader’s path traverses before that passage establishes X.
Near-miss that must not fire: a presupposition licensed by a declared prerequisite, or by an earlier passage on that same path.
Citation: quote the passage, name the path, and list what that path traverses before it — the group A rule for a negative finding.
Standing: advisory, not measured. Reported; does not block.

The repair is to supply the content. Replacing the presupposition with a pointer moves the defect into A2 rather than fixing it.

A2 — Bare cross-reference

Fires on: a pointer standing where the reader needs the content now — they cannot complete the decision the passage asks of them without leaving the document.
Near-miss that must not fire: a pointer to depth genuinely outside the document’s scope, carrying a forwarding address. The discriminator is whether the reader can proceed without following it.
Citation: quote the passage, name the path, and list what that path traverses before it — the group A rule for a negative finding.
Standing: advisory, not measured. Reported; does not block.

A3 — Undefined term at first use

Fires on: a load-bearing term relied on before it is glossed, linked, or shown by example.
Near-miss that must not fire: a term the spec’s declared audience owns. The audience table names a role plus a goal, and the role decides — not the judge’s own familiarity.
Citation: quote the passage, name the path, and list what that path traverses before it — the group A rule for a negative finding.
Standing: advisory, not measured. Reported; does not block.

Group B — what misrepresents what the reader already has

Citation this group owes. Quote both passages — the one that establishes the content and the one that misrepresents it — and confirm the two locations differ before reporting.

B1 — Re-presented as new

Fires on: already-established content returning with new-information marking — an indefinite article, an existential (there is a…), a defining move — as though it were being stated for the first time.
Near-miss that must not fire: a legitimate echo — the same content restated but marked as given (a definite article, an anaphor, as above). Content restated and marked as already given does not fire this entry.
What this entry does not fire on: the recurrence itself. This entry fires on the marking, never on the fact that a claim appears twice; a claim may and should be restated on every reader path that needs it.
Citation: quote both passages and confirm the two locations differ — the group B rule.
Standing: advisory, not measured. Reported; does not block.

B2 — Term drift

Fires on: a term introduced for one kind of subject predicated of another — a verb of holding is sound of a container (a file carries a field) and unsound of an act (a matching carries the target).
Near-miss that must not fire: a second subject class the document explicitly extends the term to, at the point of extension.
Citation: quote both uses and confirm the two locations differ — the group B rule.
Standing: advisory, not measured. Reported; does not block.

Consistent term reuse is a virtue in agent-facing prose and is exactly what produces this defect in reader-facing prose: the term propagates past the subject class it was coined for, silently.

B3 — Contradiction

Fires on: two passages whose claims cannot both hold — the document establishes X reaches what Y cannot, then routes a case away from X on the grounds that only Y is available. The second passage is usually the older one.
Near-miss that must not fire: two claims scoped to different conditions, each stating its condition. Contradictory values on different targets genuinely coexist; only the same target makes them a conflict.
Citation: quote both passages and confirm the two locations differ — the group B rule.
Standing: advisory, not measured. Reported; does not block.

Group C — where the document disagrees with its own spec

Citation this group owes. Quote the passage and the spec line it disagrees with — the prerequisite, the audience row, the declared doc type, or the coverage row. A group C finding carrying no spec quote is a preference, not a defect.

C1 — Declaration mismatch

Fires on: prose relying on a sibling document the spec declares not a prerequisite, or assuming knowledge the audience table does not grant.
Near-miss that must not fire: prose that restates that sibling’s claim in full. The sibling is referenced but not relied on, so the reader is not sent anywhere.
Citation: quote the passage and the spec line — the prerequisite or audience row — that it disagrees with; the group C rule.
Standing: advisory, not measured. Reported; does not block.

C2 — Claim without mechanism

Fires on: a chain of assertions that never supplies the causal step — the reader is told that X, and that Y follows, but never why — in a document whose declared doc type is explanation.
Gated by the declared doc type: this entry fires only where the declared type is explanation. Any passage in a tutorial, how-to, or reference document is outside it entirely.
Near-miss that must not fire: a definition, a table row, or a summary recap — and any passage in a tutorial, how-to, or reference document.
Citation: quote the passage and the spec line — here, the declared doc type — that it disagrees with; the group C rule.
Standing: advisory, not measured. Reported; does not block.

C3 — Orphan claim

Fires on: a claim the document lands and then never uses — nothing later depends on it, connects to it, or pays it off.
Near-miss that must not fire: a claim that is a payoff — the north star, or a claim serving a coverage row the spec requires for its own sake.
Citation: quote the passage and the spec line — the coverage row or north star — that it disagrees with; the group C rule.
Standing: advisory, not measured. Reported; does not block.

One finding per passage

Where a single passage fires more than one entry, report one finding for that passage.

Repair subsumption selects which one: report the entry whose repair subsumes the other’s. A catalog that reports every angle on the same sentence reads as chatty, and a chatty catalog gets routed around — which costs more than the findings it dropped.

Recurrence is retracted

An earlier revision of this bar held a claim landed in two passages to be a defect. That rule is retracted — retracted, not moved to another instrument, and not softened into a judged entry. No criterion on this bar fires on recurrence.

A claim may recur freely. It may arrive on every reader path that needs it; what it may not do is arrive where the reader cannot retrieve it.

The retracted rule’s prescribed fix — replacing the second passage with a pointer — is the worse defect. A bare cross-reference guarantees the bridging cost that recurrence only risked, which is why that substitution is itself catalogued, as A2.

The entry that survived the retraction is B1, re-presented as new, and it fires on new-information marking rather than on the restatement.

Evidence — required at both instruments

A citation names where it came from, not only what it said. Each quote carries its location — heading, and line number where the artifact has them. A quote can be transcribed perfectly and attributed to the wrong passage, and the reader of the finding has no way to tell, because the words check out.

Two quotes resolving to the same location are one passage read twice. Every criterion here is a relation between passages, so a finding whose two quotes land in the same place has not found a pair. Such a finding is not reportable. The judge compares the two locations and confirms they differ before reporting.

Standing — what blocks, and what does not

An entry’s standing is whether a finding from it stops the gate. Standing is a property of the individual entry, and it changes only by calibration.

Advisory until calibrated

A catalog entry does not block until it has been run against documents this repo already accepts and documents it already considers weak, with the entry’s false-positive rate reported rather than asserted.

Once calibrated, an entry blocks on a finding that is confirmed and undefended — only. A finding the producer defended with a rationale the judge weighed and accepted does not block, and neither does an unconfirmed one.

Why the asymmetry runs that way is argued at the doc-eval model.

Standing today

Each entry’s standing is readable beside that entry above, and here as one table:

Entry	State	False-positive rate	Corpus run
A1 unresolvable presupposition	advisory	not measured	—
A2 bare cross-reference	advisory	not measured	—
A3 undefined term at first use	advisory	not measured	—
B1 re-presented as new	advisory	not measured	—
B2 term drift	advisory	not measured	—
B3 contradiction	advisory	not measured	—
C1 declaration mismatch	advisory	not measured	—
C2 claim without mechanism	advisory	not measured	—
C3 orphan claim	advisory	not measured	—

All nine entries are advisory, so the whole catalog is non-blocking today. That is the designed starting state, not an outage: the entries are reasoned rather than measured, and reasoning is exactly what calibration exists to check.

The enumeration rule is not in this table. It is inspection, and it blocks.

The judged pass runs in two passes, in order.

Pass 1 — the blind reading. It simulates a reader on one declared control-flow path. It receives exactly three things:

the document,
that declared control-flow path, and
the audience row for that path.

Nothing else. Not the catalog, not the entry names, not the spec’s coverage table, and not the deliberate-violation record.

Pass 1 is blind to the catalog. Why the first pass must be blind is argued at the doc-eval model.

Pass 2 — the scoring. It scores pass 1’s transcript against the catalog, and it is the pass that reads the deliberate-violation record.

Declaring a deliberate violation

Any expectation about prose can be violated to good effect, so the producer may mark a judged finding intentional and have the judge weigh it.

The channel is the verification.md the producer already writes for the judge — the producer-to-judge file the judge runs and never authors. The declaration goes under a ## Deliberate violations heading, one row per claim. No second artifact is created for it.
The three fields are the catalog entry, the location, and the rationale.
When the judge reads it: in the scoring pass only. Passing it to the blind reader would name the entry and the location, which is the leak the two-pass split exists to prevent.
A rationale is weighed, not obeyed. A rationale asserting only that the choice was deliberate clears nothing; it has to say what the violation buys the reader it was made for.

Calibrating an entry

Moving one entry from advisory to blocking is a procedure. It runs in five ordered steps.

Name the corpus. At least one document the repo already accepts and one it already considers weak. The team names them — a judge that picks its own corpus has chosen the evidence that suits it.
Run the judged pass unchanged — blind simulation, then scoring — over every document in the corpus. Do not shorten it for the calibration; a procedure calibrated in a cheaper mode has measured something other than what will run.
Score against expectation, per entry — see below.
Record the false-positive rate and the named corpus together, in the standing table. A rate recorded without a named corpus is not a measurement.
An entry that fired on an already-accepted document stays advisory, and the repair is to widen that entry’s near-miss.

Scoring a run

On a document the repo already accepts, every firing counts as a false positive.
On a document the repo already considers weak, a miss does not disqualify the entry — it is informative, not fatal.
An entry that fires on an already-accepted document stays advisory. The repair is to widen its near-miss, which is usually too narrow.

Calibration is per entry

Calibration is per entry, never a majority vote across the catalog.

Entries clearing together tells you nothing about another entry. The entries fail in different ways, and one well-calibrated entry does not vouch for its neighbor.
An entry that fired on neither corpus document is untested, not calibrated. No firing is no evidence, and a row moves out of advisory only with a rate and a named corpus beside it.

What neither instrument may assert

Tone, register, length, word choice, and section order are out of scope. They are not reportable, and the exclusion holds at both instruments — the enumeration rule and the catalog alike.

Evidence is what draws that line. A criterion that can be evidenced — two quoted locations, each naming where it came from, confirmed to be two different places — is assertable. A complaint that cannot be is a style opinion, and this bar has no channel for one.

Precedence — a frozen scenario outranks this bar

Where a frozen scenario requires what a criterion here would fail, the scenario wins and the bar yields. The suite was ratified at the spec gate, and a bar that could veto a frozen scenario would make the impl gate a second spec gate.

The collision is filed as an architect observation against the spec, so the scenario is fixed where it lives. It is not reported as a gate blocker.

What this page does not own

Three lookups belong to other pages and are reached by link rather than answered here:

what each of the four scenario-scoped checks verifies, and why a document needs two instruments — the doc-eval model;
what a documentation spec must contain, and must never freeze — the spec-gate Builder bar;
which agent runs this bar, which one writes the document, and the write-versus-run independence anchor — the production chain.

For what Quill is and how to register it in a project, see the Quill overview and init-quill.