Loomworks — B-9 Step 0 inspection brief — v0.1

Version. 0.1 Date. 2026-08-03 Author. Claude.ai (drafting session). Operator: Marvin Percival. Executed by. Claude Code on DUNIN7-M4, one session. Target. /Users/dunin7/loomworks-engine at main ffc29af; /Users/dunin7/loomworks at main f477c87. Both read-only. Charter. standing-notes/dunin7-standing-authorization-charter-v0_1. R-5 inspection run. Build-list item. B-9 — change request E. Status. Read-only. Establishes ground for the provenance-walk change request. Produces no code change and no change request.

Why now. Every blocker is gone. B-25 closed — authorship is true, so a walk no longer terminates at a fabricated contributor. B-29 closed — the provenance field is true at every site. W-16 was struck — version-pinned Manifestation retrieval existed all along. And the four rooms exist, so there is a surface for the walk to live in.

What B-9 promises. Click any statement in a finished report and walk backward to the person who said it. This brief exists mostly to establish whether that sentence is buildable as written — see §3.

Grounding. inspection-briefs/loomworks-walk-audit-report-v0_1 (the five-hop walk demonstrated by hand); inspection-briefs/loomworks-b25-sizing-sweeps-findings-v0_1 §5 (what the walk does and does not touch); inspection-briefs/loomworks-b6-step-0-findings-v0_1 and loomworks-cr-c-step-0-findings-v0_1 (the current surface). All predate at least one merge — verify, do not inherit.

1. The fences

Read-only in both repositories. No fixes, no branches, no commits, no npm install.

No dev server, no build, no test run, no database, no perimeter call. playground_dev is the live production database.

Both repositories are free. No change request holds either. This session still does not build.

2. The discipline clause

State only what you have read back, plus the disciplines at standing-notes/loomworks-standing-note-executable-document-versions-v0_4:

1. Name the file and line for every claim.
2. Report unread rather than closing a gap with a plausible reading.
3. Do not inherit. Every grounding document predates a merge.
4. A sweep pattern must not encode an assumption about its target — including case.
5. Confirm a target exists before treating its silence as a pass.
6. Count the thing; do not read the figure. Where a count is the finding, count it.

3. Question one — is "click any statement" buildable?

This question decides B-9's size, and everything else in this brief is secondary to it.

The walk audit demonstrated a five-hop walk between records: render → shape reference → shape → Manifestation reference and version → selected memory references → assertion at version → attributed contributor. Every hop is record-to-record. B-9's promise is different: it starts at a statement inside a rendered artifact.

Establish whether a render's content is addressable below the level of the whole render.

• What a render's stored content actually is — HTML, Markdown, a structured document, an opaque blob. Read the stored shape, not the response schema.
• Whether anything inside it carries a reference back to a source assertion. A marker, an identifier, an anchor, a data attribute, a parallel index — anything.
• How the content is produced. If a specialist generates prose from a set of selected assertions, does anything record which assertion produced which sentence?
• Whether the Shape names its selected assertions individually, and whether that selection survives into the render in any addressable form.

The two outcomes, and they are very different builds:

Per-statement linkage exists or is derivable → B-9 is a surface build over an existing structure. Roughly the size of a room.

It does not → B-9 as promised requires the render to carry provenance it does not currently carry, which is a production-pipeline change and a substrate change, not a walk. **The walk that is buildable today is render-level**: this document was built from this Shape, from this Manifestation at this version, from these assertions, by these people. That is a real and valuable thing and it is not the promise.

Report which, with anchors. Do not design across the gap. If the answer is the second, the change request decides whether to build the render-level walk, change the pipeline, or split B-9 — and that is an Operator-facing decision, not one this session takes.

4. Question two — the five hops, re-verified post-merge

The walk audit ran by hand against an older engine. Two merges have landed on the walk's own subject matter since.

For each hop, establish the route, its parameters, and its response shape:

1. Render → the Shape it was built from.
2. Shape → the Manifestation reference and version it selected from.
3. Manifestation at that version → the memory references it organized.
4. Memory reference → the assertion at its version.
5. Assertion → the attributed contributor.

For each: does a route exist, or did the audit reach it by database inspection? A hop the audit performed with a query is not a hop the surface can make.

And specifically: hop 5 now returns true authorship, since B-25 closed. Confirm the actor kinds the walk will encounter — the four kinds, and which of them a walk should render as a person who said this versus the Companion acting for a person versus an agent. The B-25 work established these carry different authority; a walk that flattens them repeats the mistake B-25 avoided.

5. Question three — the provenance field, now that it is true

provenance.wasGeneratedBy is now the real event identifier at every site.

Establish whether the walk should use it. The B-25 sizing found the demonstrated walk does not touch it — the five hops run through references and versions, not through events. So this is a live question, not a settled one:

• Does wasGeneratedBy open a hop the reference-walk cannot make — which event produced this version of this object — and is that hop useful to an Operator, or only to a debugger?
• Is there a route that reads it today, or would exposing it be new work?

Report what it makes possible. Do not recommend.

6. Question four — the surface

6.1 Where the walk would live. The four rooms exist. Establish whether any of them has an affordance that could open a walk — a clickable assertion, a render detail view, anything.

6.2 The shared contract. A walk is a sequence of reads, each of which can fail or return nothing. Establish whether ReadState can carry a walk's states, or whether a walk needs something the contract does not have — the same question B-6's brief asked about async work, which turned out to be answered by the item rather than the contract.

6.3 The vocabulary wall. The forbidden tokens a walk component would hit, and whether an adapter would need a WIRE_BOUNDARY_FILES entry. CR-2026-162 established that an adapter needing no exemption should not get one; CR-2026-163 established the converse. Which case is this?

7. What this brief does not do

• No fixes, no change request, no design.
• Does not decide whether to build the render-level walk or change the pipeline. §3 establishes which is required; the Operator decides.
• Does not recommend whether to use wasGeneratedBy. §5 reports what it makes possible.
• Does not touch the project / engagement displacement — change request D.

8. Reporting

One findings document, Markdown primary.

Path. inspection-briefs/loomworks-b9-step-0-findings-v0_1.md. Copy to ~/Downloads.

Structure: environment (both SHAs, both tree states, and an explicit statement that no server, build, test, database or perimeter call was made) · question one first and answered plainly — per-statement linkage exists, is derivable, or does not exist · questions two to four in order · corrections preserved · unread.

Filing. Charter R-1: commit and push on clean pre-flight. Append the outcome to current-status/dunin7-status-brief.

9. Claude Code kickoff

B-9 Step 0 inspection. Read-only.

Brief: loomworks-record/inspection-briefs/loomworks-b9-step-0-inspection-brief-v0_1.md
Confirm it exists, and that it is the highest version present — numeric sort.

Read the brief in full. Charter dunin7-standing-authorization-charter-v0_1 governs.

Two repositories, both READ-ONLY, both at main — the brief's header names them.

Every grounding document predates at least one merge. Re-verify carried claims
or report them unverified.

No fixes, no branches, no commits, no npm install, no dev server, no build, no
test run, no database, no perimeter call. playground_dev is the live production
database.

Section 3 decides the size of this build and everything else is secondary to it.
Answer it plainly: does a render's content carry per-statement linkage back to
source assertions, is it derivable, or does it not exist? If it does not, say so
and do not design across the gap — that is an Operator-facing decision.

File findings per section 8. Append the outcome to the status brief.

DUNIN7 — Done In Seven LLC — Miami, Florida Loomworks — B-9 Step 0 inspection brief — v0.1 — 2026-08-03 The walk between records is proven. Whether it reaches inside a document is the question.