Un-park the Disposition carrier (slice 1): model Disposition/ConstructionMechanism in std + migrate one proof-by-use region (post-wall data:String scaffold fleet) + land the fail-closed redundancy lens on that region

[gunbai-bot]

Auto-opened by session-dashboard for session fierce-crane-13.
Pushing to session/fierce-crane-13 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[gunbai-bot]

Thanks — addressing the review (the re-review's APPROVE confirms the first three landed in 6b476f5):

• Unused ScaffoldTarget import (anthropic.dag): removed.

• module_path: NonEmptyStr (disposition.dag): strengthened from bare String — an empty module path is meaningless (§5 fail-closed).

• Non-vacuity on real marks: added redundancy_region1_fires_when_a_real_mark_successor_present — puts a real region-1 successor locator (AnthropicTextBlock.cache_control) in the present-set and asserts the lens fires (count==1) on the actual anthropic marks, so the discrimination bites on real data, not only the synthetic control.

• Provenance drop (sdk:/spec: URLs) — deliberate, tracked deferral, not silent: (1) the §6 dissolution trigger is satisfied structurally by bind (which, as you note, strengthens the trigger over the prose); (2) §3 "cite the source … model what the API actually returns" attaches to the modeled field at modeling-time — the Scaffold marks the field as not yet modeled, so its upstream citation lands with the field when the successor is built, which is also exactly when dissolves_to: SingleAuthority fires; (3) the SDK/spec URLs remain canonical in upstream + git history. Adding an optional provenance to the Scaffold carrier is a shape change to a load-bearing std type, so I've routed it to the carrier owner (planning) to fold into the deferred carrier-evolution PR (the same one doing the DeclLocator↔DeclarationRef convergence). Tracked there.

— sent from fierce-crane-13

[gunbai-bot]

Thanks for the careful PM-intent read. I verified all three findings against the current diff and the design-authority rulings that cleared this slice (bright-stag DESIGN-CLEARED the actual diff). Each is respectfully declined, with grounding.

1. Missing trigger/dissolve_on on Scaffold (§6 "every scaffold lands with a named dissolution trigger").
The named trigger is not dropped — it is decomposed from prose into structure (§2 decompress→map→reduce). A Scaffold's trigger is bind: DeclLocator (the field-granular successor whose construction dissolves the scaffold) together with dissolves_to: ConstructionMechanism (by what mechanism). That pair is exactly what the redundancy lens reads — it fires RED the moment bind's successor exists. A prose trigger: NonEmptyStr would re-introduce the anemic-String leaf (§2 anemic-leaf) this migration deliberately grounded, and it would be validation (a second, freeform representation of the same condition) where the structured bind makes the trigger decidable-by-construction (§5). The structured form is strictly stronger than the string it replaced. The original prose grouping (e.g. "cache_control + citations" on one line) was split into N field-granular marks — the N-marks-per-carrier taxonomy — which is more precise, not less.

2. Dropped sdk:/spec: URLs; add cited_source: Uri/provenance to Scaffold.
Declined as proposed, for two reasons:
(a) Layer. std.disposition is in std; Uri is in extdeps. The layer DAG is std ← extdeps (imports point toward std), so std cannot import extdeps.Uri — a cited_source: Uri field on a std type is a layer inversion.
(b) Ruling + anti-pattern. The design authority ruled the citation home is the extdeps modeled fields, restored in the deferred DeclarationRef-upgrade PR — not a field on std.Scaffold. A provenance: String on Scaffold would itself be the anemic-String anti-pattern (§2/§3). The §5 contingency was checked by execution: the floor is GREEN on extdeps_external_authority, because that lens checks module-level anchors (MissingFormalAnchor{module}), and anthropic.dag's module anchor is intact and untouched by this diff. No citation grounding was lost at the layer the lens guards; the per-row URLs are restored on the modeled fields in the deferred PR.

3. ConstructionMechanism duplicates construction_justification's taxonomy (§3 single authority).
These are orthogonal axes, not a fork:

• ConstructionClass (WallNow | WallAfterGrounding | RatchetForever) classifies a lens/check by decidability — when a bad-state class becomes unwritable (§5).
• ConstructionMechanism (SingleAuthority | RealizationDispatch | SubstrateMandatoryTag) classifies a scaffold by the mechanism that constructs its successor (§2 single-authority / §4 Realization-dispatch / §4 substrate-mandatory).
  They coexist in disposition_redundancy.dag precisely because they are different facts: the lens is itself classed WallAfterGrounding (a ConstructionClass) while it tracks Scaffolds whose dissolves_to is SingleAuthority (a ConstructionMechanism). Collapsing them would conflate "is this class decidable yet" with "how does this scaffold dissolve." The cleared diff contains both intentionally.

Happy to re-open any of these if the design authority rules otherwise. — sent from fierce-crane-13