<?xml version="1.0" encoding="utf-8"?>
<entry xmlns="http://www.w3.org/2005/Atom" xml:base="https://docs.spec-kitty.ai/adr/3.x/2026-06-26-1-single-authority-seam-and-call-site-gate/">
  <id>https://docs.spec-kitty.ai/adr/3.x/2026-06-26-1-single-authority-seam-and-call-site-gate/</id>
  <title>ADR: Single-Authority Seam + Call-Site Gate for Resolution Boundaries (Phase 1)</title>
  <updated>2026-06-26T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/3.x/2026-06-26-1-single-authority-seam-and-call-site-gate/"/>
  <summary>Handle canonicalization and surface-authority choices move behind one sanctioned seam policed by an AST call-site gate; the blind primitive stays blind to avoid recursion.</summary>
  <content type="html">&lt;h2 id=&quot;context-and-problem-statement&quot;&gt;Context and Problem Statement&lt;/h2&gt;
&lt;p&gt;Logic in this codebase reaches infrastructure / resolution boundaries &lt;strong&gt;directly&lt;/strong&gt;: a
handle (&lt;code dir=&quot;auto&quot;&gt;mission_id&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;mid8&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;slug&lt;/code&gt;) is composed into a directory, and coordination-vs-primary
artifact authority is selected, at the call site. Each boundary today is governed by a
&lt;strong&gt;convention every caller must remember&lt;/strong&gt; — “canonicalize the handle before you compose it,”
“read &lt;code dir=&quot;auto&quot;&gt;meta.json&lt;/code&gt; from the primary anchor,” “write status to the coord surface but planning to
primary.” Nothing enforces the convention; correctness is a property of every caller’s memory.&lt;/p&gt;
&lt;p&gt;The convention demonstrably leaks:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;#2164 (write-leg fold divergence).&lt;/strong&gt; The “fix” for the canonicalization boundary
introduced &lt;strong&gt;three different folds&lt;/strong&gt; of the same handle→dir canonicalization — the read seam,
the write placement, and a third site each re-implemented the rule, and they diverged. The
duplication &lt;em&gt;is&lt;/em&gt; the defect: a bare &lt;code dir=&quot;auto&quot;&gt;mid8&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;slug&lt;/code&gt; handle resolves to a different dir than the
canonical &lt;code dir=&quot;auto&quot;&gt;&amp;#x3C;slug&gt;-&amp;#x3C;mid8&gt;&lt;/code&gt; wherever a caller forgets (or mis-copies) the fold.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The #2160 class (coord-vs-primary authority bypass).&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;mark-status&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;move-task&lt;/code&gt;, and
&lt;code dir=&quot;auto&quot;&gt;safe_commit&lt;/code&gt; each independently decide which surface to write, and each can write the &lt;em&gt;wrong&lt;/em&gt;
one (coord when it should be primary, or vice versa). Same shape: an infra/authority decision
made per-caller, recurring as an N+1 family.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The naive cure — fold the guard into the blind primitive &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt; — is
&lt;strong&gt;architecturally impossible&lt;/strong&gt;: the canonicalizer &lt;code dir=&quot;auto&quot;&gt;_canonicalize_bare_modern_handle&lt;/code&gt; itself calls
that primitive, so folding canonicalization into the primitive is &lt;strong&gt;infinite recursion&lt;/strong&gt;, live-reproduced
and recorded as FR-011. The primitive is handle-blind &lt;em&gt;by contract&lt;/em&gt; and must stay so. The guard
must therefore live at the &lt;strong&gt;seam&lt;/strong&gt; in front of the primitive, not inside it.&lt;/p&gt;
&lt;p&gt;This is not a foreign architecture — the repo already lives by the pattern. &lt;code dir=&quot;auto&quot;&gt;resolution.py&lt;/code&gt;’s core
is explicitly pure with I/O threaded in by an imperative shell; &lt;code dir=&quot;auto&quot;&gt;RuntimeEventEmitter(Protocol)&lt;/code&gt; +
&lt;code dir=&quot;auto&quot;&gt;NullEmitter&lt;/code&gt; is a shipped port seam; and — most relevantly — &lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt;
already enforces a single-authority free function (&lt;code dir=&quot;auto&quot;&gt;ProtectionPolicy.resolve&lt;/code&gt;) with an AST
call-site gate and a curated allowlist. #2173 Phase 1 &lt;strong&gt;generalizes that existing idiom&lt;/strong&gt; to the
resolution/placement boundaries, rather than importing a new mechanism.&lt;/p&gt;
&lt;h2 id=&quot;decision-drivers&quot;&gt;Decision Drivers&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Close the bug class by construction&lt;/strong&gt;, not by per-caller memory — a forgotten or mis-copied fold must be a CI failure, not a latent runtime divergence.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Minimal churn, no partial-adoption tax&lt;/strong&gt; — the remedy must not split the codebase into “ported” and “un-ported” halves with two coexisting conventions.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Match an idiom the repo already uses&lt;/strong&gt; — &lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt; is the proven, shipped precedent; copy it rather than invent.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Preserve no-silent-fallback and fail-closed semantics&lt;/strong&gt; — ambiguity raises a typed error (&lt;code dir=&quot;auto&quot;&gt;MissionSelectorAmbiguous&lt;/code&gt;, C-009 / WP07); a cold-miss fails loud, never a verbatim passthrough.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Stay out of the blind primitive&lt;/strong&gt; — folding the guard into &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt; recurses (FR-011); the guard lives at the seam.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;considered-options&quot;&gt;Considered Options&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Status-quo per-caller canonicalization convention&lt;/strong&gt; — every caller remembers to canonicalize / pick the right surface.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Fold the guard into the blind primitive&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Single sanctioned seam + AST call-site gate&lt;/strong&gt; (generalize &lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt;) — &lt;strong&gt;chosen for Phase 1&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Full injectable DI port&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;MissionResolver&lt;/code&gt; Protocol + real/fake adapters) — &lt;strong&gt;deferred to Phase 2&lt;/strong&gt;.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;decision-outcome&quot;&gt;Decision Outcome&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Chosen option:&lt;/strong&gt; “Single sanctioned seam + AST call-site gate” (Option 3), because it closes the #2164 / #2160 bug class &lt;strong&gt;by construction&lt;/strong&gt; at roughly a tenth of the churn of a full DI port, with
no partial-adoption tax, by generalizing a pattern the repository already ships and trusts.&lt;/p&gt;
&lt;p&gt;Every resolution/placement boundary &lt;strong&gt;delegates to the single canonical fold&lt;/strong&gt; rather than
re-implementing it:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The handle→dir canonicalization boundary (#2164) routes through &lt;strong&gt;one&lt;/strong&gt; canonicalizer; no caller composes a raw handle.&lt;/li&gt;
&lt;li&gt;The coord-vs-primary authority boundary (#2160) routes through &lt;strong&gt;one&lt;/strong&gt; authority decision; &lt;code dir=&quot;auto&quot;&gt;mark-status&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;move-task&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;safe_commit&lt;/code&gt; do not each pick a surface.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;An &lt;strong&gt;AST architectural gate&lt;/strong&gt;, modeled directly on &lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt;, makes a
bypass a CI failure: a raw &lt;code dir=&quot;auto&quot;&gt;KITTY_SPECS_DIR / &amp;#x3C;unresolved handle&gt;&lt;/code&gt; compose, or a coord-vs-primary
authority decision made outside the sanctioned seam, fails the gate. A &lt;strong&gt;curated allowlist&lt;/strong&gt; names
the one sanctioned authority per boundary, exactly as the FR-010 gate names
&lt;code dir=&quot;auto&quot;&gt;protection_policy.py&lt;/code&gt; + &lt;code dir=&quot;auto&quot;&gt;commit_helpers.py&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Non-negotiables (binding across all squads):&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The guard lives at the &lt;strong&gt;seam&lt;/strong&gt;; the primitive &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt; stays &lt;strong&gt;blind&lt;/strong&gt; (folding it in recurses — FR-011).&lt;/li&gt;
&lt;li&gt;Ambiguity raises &lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;MissionSelectorAmbiguous&lt;/code&gt;&lt;/strong&gt; (C-009 / WP07) — &lt;strong&gt;never&lt;/strong&gt; a silent pick.&lt;/li&gt;
&lt;li&gt;A cold-miss (handle absent from the resolver) is &lt;strong&gt;fail-closed-loud&lt;/strong&gt; — never verbatim passthrough.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;One&lt;/strong&gt; sanctioned authority per boundary (no parallel resolvers); the allowlist is the explicit, reviewed record of it.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;consequences&quot;&gt;Consequences&lt;/h3&gt;
&lt;h4 id=&quot;positive&quot;&gt;Positive&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;By-construction closure&lt;/strong&gt; of the #2164 (canonicalizer) and #2160 (coord-authority) classes: a future caller cannot silently re-introduce a raw compose or a wrong-surface write — the gate fails CI.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;~10% of the port’s churn&lt;/strong&gt; — one delegation per seam + one convergence test + one AST gate, versus a Protocol + adapter + stub + DI seam threaded through ~80 call sites with ~27 test rewrites.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No partial-adoption tax&lt;/strong&gt; — there is no “ported vs un-ported” half-state; a site either delegates to the sanctioned fold or trips the gate.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Generalizes a proven in-repo pattern&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt;) — no new mechanism, no framework, no god-object.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Applies uniformly&lt;/strong&gt; to both instances: the same seam-plus-gate shape fences the canonicalizer (#2164) and the coord-vs-primary authority (#2160).&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;The gate does NOT consolidate the duplicated &lt;code dir=&quot;auto&quot;&gt;kitty-specs/&lt;/code&gt; enumeration&lt;/strong&gt; (3+ parallel walkers) or the per-call uncached re-walk / TOCTOU — that genuine consolidation needs the &lt;strong&gt;Phase 2&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;MissionResolver&lt;/code&gt; port and is explicitly out of scope here.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;AST gates carry their own maintenance&lt;/strong&gt; — line-pin / allowlist upkeep, and the curated sanctioned-seam allowlist must be deliberately extended (with a rationale comment naming the one flow each entry authorises), per the FR-010 precedent.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Forward-only&lt;/strong&gt; — the gate fences &lt;em&gt;new&lt;/em&gt; bypasses; it does not retroactively reconcile a mission already written to the wrong surface (the flatten / manual-recovery flow remains the remedy for legacy splits).&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;Status/coord write destinations are unchanged — this decision governs &lt;em&gt;how&lt;/em&gt; a boundary is crossed (one sanctioned seam), not &lt;em&gt;which&lt;/em&gt; surface a given artifact kind targets (the kind partition, settled by the predecessor ADRs).&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;confirmation&quot;&gt;Confirmation&lt;/h3&gt;
&lt;p&gt;The decision is confirmed when: (1) a parametrized convergence test proves the read-seam and
write-seam resolve &lt;strong&gt;identically for every handle form&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;mission_id&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;mid8&lt;/code&gt;, numeric, &lt;code dir=&quot;auto&quot;&gt;slug&lt;/code&gt;);
(2) the AST gate, with a runnable synthetic self-test, &lt;strong&gt;fails&lt;/strong&gt; on a deliberately re-introduced raw
compose / wrong-surface write; and (3) #2164 and the #2160 class no longer reproduce against a
&lt;strong&gt;live coordination-topology mission that genuinely diverges&lt;/strong&gt; from primary (a stubbed/flattened
fixture reproduces the false-green and is rejected). Confidence: high for bug-class closure — the
pattern is already shipped and trusted for the protected-branch boundary.&lt;/p&gt;
&lt;h2 id=&quot;pros-and-cons-of-the-options&quot;&gt;Pros and Cons of the Options&lt;/h2&gt;
&lt;h3 id=&quot;option-1--status-quo-per-caller-canonicalization-convention&quot;&gt;Option 1 — Status-quo per-caller canonicalization convention&lt;/h3&gt;
&lt;p&gt;Each caller remembers to canonicalize the handle / select the surface before composing.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Zero new infrastructure; the code already works for canonical handles.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;#2164 proves it leaks&lt;/strong&gt; — the convention propagated &lt;strong&gt;three variant folds&lt;/strong&gt; that diverged.&lt;/li&gt;
&lt;li&gt;The #2160 class is the same leak on the authority axis (mark-status / move-task / safe_commit).&lt;/li&gt;
&lt;li&gt;Correctness depends on every caller’s memory; the N+1 returns on the next edit.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-2--fold-the-guard-into-the-blind-primitive&quot;&gt;Option 2 — Fold the guard into the blind primitive&lt;/h3&gt;
&lt;p&gt;Make &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt; canonicalize internally.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Conceptually “one place,” if it could work.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Infinite recursion (FR-011), live-reproduced&lt;/strong&gt; — &lt;code dir=&quot;auto&quot;&gt;_canonicalize_bare_modern_handle&lt;/code&gt; calls the primitive, so folding canonicalization into it recurses.&lt;/li&gt;
&lt;li&gt;Breaks the primitive’s deliberate handle-blind contract.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-3--single-sanctioned-seam--ast-call-site-gate-chosen&quot;&gt;Option 3 — Single sanctioned seam + AST call-site gate (CHOSEN)&lt;/h3&gt;
&lt;p&gt;Every boundary delegates to one canonical fold; an AST gate + curated allowlist make a bypass a CI failure (generalizing &lt;code dir=&quot;auto&quot;&gt;test_protection_resolver_call_sites.py&lt;/code&gt;).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Closes the class &lt;strong&gt;by construction&lt;/strong&gt;; ~10% of the port’s churn; no partial-adoption tax; generalizes a proven idiom; applies uniformly to #2164 and #2160; preserves typed-ambiguity + fail-closed.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Does not consolidate enumeration / TOCTOU (Phase 2); AST gate carries line-pin / allowlist maintenance.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-4--full-injectable-di-port-deferred-to-phase-2&quot;&gt;Option 4 — Full injectable DI port (DEFERRED to Phase 2)&lt;/h3&gt;
&lt;p&gt;A &lt;code dir=&quot;auto&quot;&gt;MissionResolver&lt;/code&gt; Protocol + &lt;code dir=&quot;auto&quot;&gt;FsMissionResolver&lt;/code&gt; + &lt;code dir=&quot;auto&quot;&gt;FakeMissionResolver&lt;/code&gt;, owning the one
&lt;code dir=&quot;auto&quot;&gt;kitty-specs/&lt;/code&gt; walk, threaded into &lt;code dir=&quot;auto&quot;&gt;resolution.py&lt;/code&gt;’s shell via &lt;code dir=&quot;auto&quot;&gt;resolver=None&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Consolidates the 3+ parallel walkers, the per-call re-walk, and the TOCTOU; makes the pure builder FS-free-testable; is the concrete first move on #1619’s spine.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Over-built for the #2164 / #2160 bug-closure alone&lt;/strong&gt; — &gt;10× the diff (Protocol + adapter + stub + DI seam across ~80 sites + ~27 test rewrites) to close the &lt;em&gt;same&lt;/em&gt; class the gate closes.&lt;/li&gt;
&lt;li&gt;It is the &lt;strong&gt;enumeration-consolidation / #1619-unblock&lt;/strong&gt; layer, a distinct strategic concern — recorded here as the &lt;strong&gt;deferred follow-on&lt;/strong&gt;, not this decision.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Ratification (2026-07-29, write-side-seam-matrix-tracer-01KYP3MH WP02, HiC-approved).&lt;/strong&gt;
This ADR was &lt;code dir=&quot;auto&quot;&gt;status: Proposed&lt;/code&gt; but de-facto shipped — the coord-authority gate it
governs (&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_resolution_authority_gates.py&lt;/code&gt;) has been live and
enforced since the sibling &lt;code dir=&quot;auto&quot;&gt;single-authority-resolution-gates-01KW1P0F&lt;/code&gt; mission. WP02
re-pins the gate’s census floor (&lt;code dir=&quot;auto&quot;&gt;COORD_AUTHORITY_WRITE_FLOOR&lt;/code&gt; 4 → 3, “Move A” per
&lt;a href=&quot;../../../kitty-specs/write-side-seam-matrix-tracer-01KYP3MH/contracts/coord-authority-gate.md&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;contracts/coord-authority-gate.md&lt;/code&gt;&lt;/a&gt;)
by routing &lt;code dir=&quot;auto&quot;&gt;decisions/emit.py&lt;/code&gt; off the kind-blind resolver onto the kind-aware
&lt;code dir=&quot;auto&quot;&gt;placement_seam(...).read_dir(STATUS_STATE)&lt;/code&gt; seam — the natural site to formally ratify
this ADR to &lt;code dir=&quot;auto&quot;&gt;Accepted&lt;/code&gt;. A future &lt;strong&gt;Move B&lt;/strong&gt; — recognizing &lt;code dir=&quot;auto&quot;&gt;write_target(kind=...)&lt;/code&gt; calls
as sanctioned coord-write authority in the gate’s predicate — is named here as the
&lt;strong&gt;amendment target&lt;/strong&gt; for this ADR: it changes the gate mechanism itself (not merely the
census), so it must amend this ADR (def-use gated, with a preserved non-vacuity proof),
never land as a contract-only predicate widen.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Amendment (2026-09-28, &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/5146&quot;&gt;#5146&lt;/a&gt;):&lt;/strong&gt;
&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_resolution_authority_gates.py&lt;/code&gt; was deleted in #3285 (&lt;code dir=&quot;auto&quot;&gt;177e06269&lt;/code&gt;), and the
coord-authority write floor (&lt;code dir=&quot;auto&quot;&gt;COORD_AUTHORITY_WRITE_FLOOR&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;_COORD_WRITE_BY_DESIGN&lt;/code&gt;) was retired with
it, so the Move B amendment target above no longer has a gate to amend. The call-site guarantee over
&lt;code dir=&quot;auto&quot;&gt;resolve_feature_dir_for_mission&lt;/code&gt; is now enforced by the census in
&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_no_read_side_bypass.py&lt;/code&gt;; the write side is covered by
&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_no_write_side_rederivation.py&lt;/code&gt; and
&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_write_surface_placement_guard.py&lt;/code&gt;. The def-use handle-canonicality check has no
successor: its remaining target, the public &lt;code dir=&quot;auto&quot;&gt;primary_feature_dir_for_mission&lt;/code&gt; wrapper, was itself deleted
(T035).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Phase 2 (deferred follow-on, under &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1619&quot;&gt;#1619&lt;/a&gt;).&lt;/strong&gt;
The &lt;code dir=&quot;auto&quot;&gt;MissionResolver&lt;/code&gt; port — Protocol contract + one real FS adapter + one fake, owning the single
&lt;code dir=&quot;auto&quot;&gt;kitty-specs/&lt;/code&gt; walk and threaded into the resolution shell via the existing default-param idiom
(&lt;code dir=&quot;auto&quot;&gt;resolver: Port | None = None&lt;/code&gt; → &lt;code dir=&quot;auto&quot;&gt;resolver or RealAdapter()&lt;/code&gt;) — is the strategic enumeration-consolidation
layer that this Phase 1 gate deliberately does &lt;strong&gt;not&lt;/strong&gt; attempt. It kills the multi-walker duplication,
the per-call uncached re-walk, and the TOCTOU, and makes the pure builder FS-free-testable (the #1619
unblock). It is the layer the operator’s original resolver-port thesis pointed at, sharpened: ports inject
into the &lt;strong&gt;shell / builder&lt;/strong&gt;, never onto the frozen &lt;code dir=&quot;auto&quot;&gt;MissionExecutionContext&lt;/code&gt; value object. It is recorded
here as the next strategic move, not part of this decision.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Synthesis / investigation record: &lt;a href=&quot;../../../docs/plans/engineering-notes/2173-infra-logic-separation/00-SYNTHESIS.md&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;docs/plans/engineering-notes/2173-infra-logic-separation/00-SYNTHESIS.md&lt;/code&gt;&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Cross-references: &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2173&quot;&gt;#2173&lt;/a&gt;, &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1619&quot;&gt;#1619&lt;/a&gt;, &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2164&quot;&gt;#2164&lt;/a&gt;, &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2160&quot;&gt;#2160&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Predecessor ADRs: &lt;a href=&quot;/adr/3.x/2026-06-24-1-kind-and-topology-aware-artifact-placement/&quot;&gt;2026-06-24-1 — Kind- and Topology-Aware Artifact Placement&lt;/a&gt; · &lt;a href=&quot;/adr/3.x/2026-06-22-1-mission-topology-ssot/&quot;&gt;2026-06-22-1 — MissionTopology SSOT&lt;/a&gt; · &lt;a href=&quot;/adr/3.x/2026-06-24-2-write-branch-resolution-primary-anchor/&quot;&gt;2026-06-24-2 — Write-Branch Resolution PRIMARY Anchor&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Gate precedent to copy: &lt;a href=&quot;../../../tests/architectural/test_protection_resolver_call_sites.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_protection_resolver_call_sites.py&lt;/code&gt;&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
