DUNIN7-M4 — INFRASTRUCTURE CHANGE REQUEST

CR-2026-182 — B-17 (portfolio filter) and B-43 (walk the other way) (v0.1)

Version. 0.1 · Date. 2026-08-06 · Author. Claude.ai (drafting) / Marvin Percival (approving). Target. /Users/dunin7/loomworks-engine and /Users/dunin7/loomworks — CC confirms which each item touches at Step 1. Baseline. Post CR-2026-181. CC confirms exact HEAD at Step 0. Companion to. standing-notes/dunin7-build-list-v0_29 (B-17, B-43); B-9's closure (the provenance walk, which B-43 inverts). Charter. Autonomous regime. Status. Ready for CC execution, with scoped reads first.

1. Summary

Two items paired because both are described in the build list in one line each, and Claude.ai does not hold enough detail on either to write a specification that wouldn't be invention. Both get a read step inside this CR rather than a guessed §2.

B-17 — the portfolio filter. READY in the build list, no further description. A filter across engagements, presumably.
B-43 — walk the other way. READY — costed. The build list says a cost estimate exists; CC locates it rather than re-deriving. B-9 built the walk from a finished statement back to its author. B-43 is the inverse direction.

If Step 1 finds either is larger than the build list implies, it leaves this CR and gets its own scoping note. Pairing them is a convenience, not a commitment to doing both together.

2. Scope

Step 1 — read both (read-only).

B-17: find any scoping or design material on the portfolio filter in the record. Determine what "portfolio" means here in project terms (a set of engagements? something else?), what the filter filters on, and where it belongs. If nothing beyond the one-line build-list entry exists anywhere, say so — that means this is a scoping item, not a build item, and it halts.
B-43: locate the existing cost estimate. Determine what the inverse walk actually returns — from an origin, forward to everything derived from it. Confirm whether B-9's machinery is reusable in reverse or whether this needs its own traversal.

Report both before building either.

Step 2 — build whichever Step 1 confirmed as build-ready. If only one qualifies, build that one and report the other as leaving the CR. This is expected and is not a failure.

Step 3 — tests appropriate to whichever landed.

3. Governing constraints (both items)

Only show what is available. No disabled controls, no grayed options. If a filter has nothing to filter, it doesn't appear.
Plain terms on operator-facing surfaces. Memory, Manifestation, Shaping, Rendering, engagement, assertion — in plain terms, not codebase shorthand.
B-43 specifically: the walk shows provenance. Anything it displays must be traceable, not inferred-and-presented-as-fact — the same posture B-9 holds.

4. Build steps

| Step | What | Mode | |---|---|---| | 0 | Pre-flight. | Auto | | 1 | Read both. Report. Halt on either if it's a scoping item, not a build item. | Auto, conditional halt | | 2 | Build B-17 if confirmed. | Auto | | 3 | Build B-43 if confirmed. | Auto | | 4 | Tests for whatever landed. | Auto | | 5 | Full suite (+ surface sweep if touched); ruff 0; mypy at/below baseline. | Auto | | A | Checkpoint. | Checkpoint | | 6 | Tag cr-2026-182-b17-b43. Push. | Auto |

5. Acceptance gate

Step 1's reads reported for both items regardless of what got built.
Whatever built has tests and passes the full suite.
Anything that left the CR is named clearly, with why, so it can be scoped rather than lost.
Both gates clean.

6. Kickoff prompt


Execute CR-2026-182 v0.1 at ~/Downloads/cr-2026-182-loomworks-b17-b43-
v0_1.md. Confirm the CR number against the ledger first.

Two items, paired only because Claude.ai doesn't hold enough detail on
either to write a real specification without inventing one. Step 1 is
a read for both.

B-17 (portfolio filter): find any scoping/design material in the record.
If nothing beyond the one-line build-list entry exists anywhere, say so
and HALT on it — that makes it a scoping item, not a build item.

B-43 (walk the other way): the build list says it's "costed" — locate
that estimate rather than re-deriving it. Determine whether B-9's walk
machinery is reusable in reverse or whether this needs its own traversal.

It is fine and expected for only one of the two to survive Step 1.
Build what qualifies, report what leaves, don't force both.

Baseline: post CR-2026-181. Confirm current HEAD at Step 0.

DUNIN7 — Done In Seven LLC — Miami, Florida CR-2026-182 — B-17 and B-43 — v0.1 — 2026-08-06