<?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/2026-08-02-1-charter-wheel-assessment/">
  <id>https://docs.spec-kitty.ai/adr/3.x/2026-08-02-1-charter-wheel-assessment/</id>
  <title>ADR: Charter-Wheel Assessment — Extractable in Principle, Cut Over Kernel→Doctrine→Charter as One No-Partial Follow-On</title>
  <updated>2026-08-02T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/3.x/2026-08-02-1-charter-wheel-assessment/"/>
  <summary>`src/charter` carries zero `specify_cli` import edges and is extractable in principle, but the kernel, doctrine and charter wheels must cut over together, never in stages.</summary>
  <content type="html">&lt;h2 id=&quot;context-and-problem-statement&quot;&gt;Context and Problem Statement&lt;/h2&gt;
&lt;p&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3101&quot;&gt;#3101&lt;/a&gt; wants &lt;code dir=&quot;auto&quot;&gt;src/doctrine/&lt;/code&gt; — and,
assessed here, &lt;code dir=&quot;auto&quot;&gt;src/charter/&lt;/code&gt; — to ship as installable wheels on the documented dependency
chain, shown below, which runs from the dependency-free kernel at the root through doctrine
and charter to glossary/runtime and finally &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt;.&lt;/p&gt;
&lt;div class=&quot;expressive-code&quot;&gt;&lt;figure class=&quot;frame not-content&quot;&gt;&lt;figcaption class=&quot;header&quot;&gt;&lt;/figcaption&gt;&lt;pre data-language=&quot;plaintext&quot;&gt;&lt;code&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;kernel (root) &amp;#x3C;- doctrine &amp;#x3C;- charter &amp;#x3C;- glossary/runtime &amp;#x3C;- specify_cli&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/figure&gt;&lt;/div&gt;
&lt;p&gt;The Mission &lt;code dir=&quot;auto&quot;&gt;doctrine-charter-split-unification-01KZ0SRB&lt;/code&gt; was scoped as &lt;strong&gt;groundwork for&lt;/strong&gt;
that cutover, not the cutover (spec C-002). Two of its work packages changed what is
knowable about charter’s extractability, and this ADR records the resulting assessment plus
the sequencing decision, so the follow-on is mechanical rather than exploratory:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;WP10 / FR-008&lt;/strong&gt; deleted the single real upward layer edge out of the charter layer — a
lazy in-function &lt;code dir=&quot;auto&quot;&gt;import specify_cli&lt;/code&gt; in &lt;code dir=&quot;auto&quot;&gt;charter/synthesizer/synthesize_pipeline.py&lt;/code&gt;,
used only as a &lt;code dir=&quot;auto&quot;&gt;__version__&lt;/code&gt; fallback — and shipped a &lt;strong&gt;non-vacuous AST-walk gate&lt;/strong&gt; for
it. The pre-existing &lt;code dir=&quot;auto&quot;&gt;pytestarch&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;LayerRule&lt;/code&gt; in
&lt;a href=&quot;../../../tests/architectural/test_layer_rules.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_layer_rules.py&lt;/code&gt;&lt;/a&gt;
was &lt;strong&gt;green with the violation present&lt;/strong&gt;, because pytestarch’s import resolution
(&lt;code dir=&quot;auto&quot;&gt;ImportConverter.convert&lt;/code&gt;) recurses into a module’s and a function’s &lt;code dir=&quot;auto&quot;&gt;.body&lt;/code&gt; but never
into an &lt;code dir=&quot;auto&quot;&gt;except&lt;/code&gt; handler’s body — and the deleted edge sat inside an &lt;code dir=&quot;auto&quot;&gt;except Exception:&lt;/code&gt;
block, so it was invisible to that specific blind spot, not to function-scope imports in
general. A gate that cannot fail on the one violation it nominally guards is worse than no
gate: it manufactures false confidence.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;WP12 / FR-009+FR-010&lt;/strong&gt; minted &lt;a href=&quot;../../../src/kernel/pyproject.toml&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;src/kernel/pyproject.toml&lt;/code&gt;&lt;/a&gt;
(&lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt;, zero first-party dependencies — the true root of the chain) and
closed &lt;code dir=&quot;auto&quot;&gt;src/doctrine/pyproject.toml&lt;/code&gt; with its real
&lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt; dependency plus a working hatchling build hook for the &lt;strong&gt;out-of-tree&lt;/strong&gt;
&lt;code dir=&quot;auto&quot;&gt;packs/&lt;/code&gt; tree. Both were &lt;strong&gt;build-verified with a real &lt;code dir=&quot;auto&quot;&gt;hatch build&lt;/code&gt;&lt;/strong&gt;, not asserted from
manifest shape.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The open architectural question this ADR answers: &lt;em&gt;given a charter layer with no upward
import edge, can &lt;code dir=&quot;auto&quot;&gt;src/charter&lt;/code&gt; simply be lifted into its own wheel — and if not now, in what
order and under what discipline?&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;The failure mode to avoid is already documented in this repository. ADR
&lt;a href=&quot;/adr/3.x/2026-04-25-1-shared-package-boundary/&quot;&gt;2026-04-25-1&lt;/a&gt; records
&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/pull/779&quot;&gt;PR #779&lt;/a&gt;: a &lt;em&gt;partial&lt;/em&gt; package cutover
that moved code into the target tree while leaving the old production imports alive. The
resulting hybrid was structurally identical to the pre-cutover state from a clean-install
perspective, and it re-imposed cross-package release lockstep. It was rejected.&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;Claim precision.&lt;/strong&gt; An over-broad “charter is decoupled” claim would be load-bearing for
the follow-on and is not what the landed evidence supports. The claim must be no wider
than its proof.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Transitive closure, not local cleanliness.&lt;/strong&gt; Charter imports &lt;code dir=&quot;auto&quot;&gt;doctrine&lt;/code&gt; (109 import
statements) and &lt;code dir=&quot;auto&quot;&gt;kernel&lt;/code&gt; (8). A charter wheel with no doctrine wheel and no kernel wheel
beneath it is unresolvable at install time, regardless of how clean charter itself is.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No-partial discipline.&lt;/strong&gt; The #779 precedent makes a half-executed cutover the specific
named risk, not a hypothetical one.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Reuse the landed boundary pattern.&lt;/strong&gt; This repository already has an enforcement triad
that works. Inventing a second, parallel packaging-enforcement idiom would split the
authority and rot.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;C-002 must hold.&lt;/strong&gt; The groundwork Mission must not leave the tree in a state where CI
builds or installs a nested wheel whose &lt;code dir=&quot;auto&quot;&gt;kernel&lt;/code&gt; dependency is not yet published.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;considered-options&quot;&gt;Considered Options&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Option A — Extract the charter wheel now&lt;/strong&gt;, on top of the landed WP10/WP12 groundwork.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option B — Assess and sequence: one atomic kernel→doctrine→charter cutover, deferred&lt;/strong&gt;
as an explicit no-partial follow-on that extends the &lt;code dir=&quot;auto&quot;&gt;2026-04-25-1&lt;/code&gt; pattern.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option C — Abandon the split&lt;/strong&gt;; keep kernel/doctrine/charter permanently in the root
wheel and close #3101.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;decision-outcome&quot;&gt;Decision Outcome&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Chosen option: “Option B”&lt;/strong&gt;, because charter is extractable &lt;em&gt;in principle&lt;/em&gt; but not &lt;em&gt;in
isolation&lt;/em&gt;: its transitive closure needs the kernel and doctrine wheels published first, and
splitting the move into stages is precisely the partial cutover the #779 precedent forbids.&lt;/p&gt;
&lt;h3 id=&quot;a-the-extractability-claim-scoped-to-its-proof&quot;&gt;(a) The extractability claim, scoped to its proof&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;src/charter&lt;/code&gt; is extractable in principle: it carries zero &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; import
entanglement — that is, zero static &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; import edges at any scope, module-level or
in-function — as proven by WP10’s AST-walk gate,
&lt;a href=&quot;../../../tests/architectural/test_charter_no_specify_cli_import.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_charter_no_specify_cli_import.py&lt;/code&gt;&lt;/a&gt;.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;That gate is the durable proof and the exact boundary of the claim. It walks the full AST
(&lt;code dir=&quot;auto&quot;&gt;ast.walk&lt;/code&gt;) of every module under &lt;code dir=&quot;auto&quot;&gt;src/charter/**&lt;/code&gt;, so in-function, in-class, in-&lt;code dir=&quot;auto&quot;&gt;try&lt;/code&gt;, and
in-&lt;code dir=&quot;auto&quot;&gt;except&lt;/code&gt;-handler imports are all caught — including the exact &lt;code dir=&quot;auto&quot;&gt;except&lt;/code&gt; blind spot
&lt;code dir=&quot;auto&quot;&gt;pytestarch&lt;/code&gt;’s converter misses. Its
non-vacuity was demonstrated by self-mutation (NFR-004): re-adding the deleted in-function
import turns it red naming the exact line, while &lt;code dir=&quot;auto&quot;&gt;test_layer_rules&lt;/code&gt; stays green throughout.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;What this gate does NOT prove.&lt;/strong&gt; Citing it for anything below is a misreading:&lt;/p&gt;





























&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Not covered&lt;/th&gt;&lt;th&gt;Why it is out of scope&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;importlib.import_module(...)&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;__import__(...)&lt;/code&gt; string indirection&lt;/td&gt;&lt;td&gt;The gate matches &lt;code dir=&quot;auto&quot;&gt;ast.Import&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;ast.ImportFrom&lt;/code&gt; nodes; a name assembled as a string is invisible to it. (Charter has no such call site today — but that is an observation, not an enforced invariant.)&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Entry points / plugin discovery&lt;/td&gt;&lt;td&gt;Resolved at runtime from installed distribution metadata; never an AST edge.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Duck-typed objects passed in as data&lt;/td&gt;&lt;td&gt;A caller handing charter a &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt;-owned object is a real runtime coupling with no import edge anywhere in &lt;code dir=&quot;auto&quot;&gt;src/charter/**&lt;/code&gt;.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Non-import couplings — shared on-disk paths, config keys, environment contracts&lt;/td&gt;&lt;td&gt;Not imports.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;Charter ↔ glossary / runtime / re-export edges&lt;/strong&gt;&lt;/td&gt;&lt;td&gt;Explicitly &lt;strong&gt;unproven&lt;/strong&gt; by this gate (research.md D10). The gate’s forbidden root is &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; alone. These edges are a separate assessment the cutover follow-on must perform.&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;So: “zero &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; import entanglement”. Not “zero entanglement”.&lt;/p&gt;
&lt;h3 id=&quot;b-the-cutover-is-sequenced-kerneldoctrinecharter-and-is-no-partial&quot;&gt;(b) The cutover is sequenced kernel→doctrine→charter, and is NO-PARTIAL&lt;/h3&gt;
&lt;p&gt;The publish order is forced by the dependency chain — &lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt; has zero
first-party dependencies, &lt;code dir=&quot;auto&quot;&gt;spec-kitty-doctrine&lt;/code&gt; depends on kernel, a future
&lt;code dir=&quot;auto&quot;&gt;spec-kitty-charter&lt;/code&gt; depends on both:&lt;/p&gt;
&lt;div class=&quot;expressive-code&quot;&gt;&lt;figure class=&quot;frame not-content&quot;&gt;&lt;figcaption class=&quot;header&quot;&gt;&lt;/figcaption&gt;&lt;pre data-language=&quot;plaintext&quot;&gt;&lt;code&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;spec-kitty-kernel  -&gt;  spec-kitty-doctrine  -&gt;  spec-kitty-charter&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;   &lt;/span&gt;&lt;/span&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;(zero first-party deps)     (needs kernel)        (needs kernel + doctrine)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/figure&gt;&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;This ordering is a publish sequence inside one cutover, not a schedule of three shippable
milestones.&lt;/strong&gt; A partial cutover is forbidden: ADR
&lt;a href=&quot;/adr/3.x/2026-04-25-1-shared-package-boundary/&quot;&gt;2026-04-25-1&lt;/a&gt; records, under &lt;em&gt;Alternatives
considered&lt;/em&gt;, that constraint &lt;strong&gt;C-007&lt;/strong&gt; of the shared-package-boundary Mission spec
explicitly forbids partial cutovers, with PR #779 as the cautionary example. The follow-on
therefore lands the three wheels together — root-wheel &lt;code dir=&quot;auto&quot;&gt;packages&lt;/code&gt; removal, published
dependency declarations, boundary enforcement, and clean-install proof in one reviewed unit
— or it does not land. Stopping after kernel+doctrine, with charter still bundled in the
root wheel while its dependencies are external, reproduces the #779 hybrid exactly.&lt;/p&gt;
&lt;h3 id=&quot;c-this-adr-extends-2026-04-25-1-it-does-not-introduce-a-new-pattern&quot;&gt;(c) This ADR extends &lt;code dir=&quot;auto&quot;&gt;2026-04-25-1&lt;/code&gt;; it does not introduce a new pattern&lt;/h3&gt;
&lt;p&gt;The follow-on reuses the existing, landed enforcement triad. Each mechanism gets a charter
analogue — same idiom, same file conventions, no second authority:&lt;/p&gt;

























&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;&lt;code dir=&quot;auto&quot;&gt;2026-04-25-1&lt;/code&gt; mechanism&lt;/th&gt;&lt;th&gt;Landed instance&lt;/th&gt;&lt;th&gt;Charter-cutover analogue&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;Boundary test&lt;/strong&gt; (no forbidden import edge, pytestarch + AST fallback)&lt;/td&gt;&lt;td&gt;&lt;a href=&quot;../../../tests/architectural/test_shared_package_boundary.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_shared_package_boundary.py&lt;/code&gt;&lt;/a&gt;&lt;/td&gt;&lt;td&gt;&lt;strong&gt;Already landed&lt;/strong&gt; as &lt;a href=&quot;../../../tests/architectural/test_charter_no_specify_cli_import.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;test_charter_no_specify_cli_import.py&lt;/code&gt;&lt;/a&gt; (WP10). The follow-on extends its forbidden-root set to cover glossary/runtime once those edges are assessed.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;pyproject-shape test&lt;/strong&gt; (no exact pins, no committed path/editable sources, no retired deps)&lt;/td&gt;&lt;td&gt;&lt;a href=&quot;../../../tests/architectural/test_pyproject_shape.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_pyproject_shape.py&lt;/code&gt;&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Extend with the same assertions for &lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;spec-kitty-doctrine&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;spec-kitty-charter&lt;/code&gt;: compatibility ranges in pyproject, exact pins only in &lt;code dir=&quot;auto&quot;&gt;uv.lock&lt;/code&gt;, no &lt;code dir=&quot;auto&quot;&gt;[tool.uv.sources]&lt;/code&gt; path entries. Partly pre-figured by WP12’s &lt;code dir=&quot;auto&quot;&gt;test_doctrine_wheel_closure.py&lt;/code&gt;.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;clean-install-verification&lt;/code&gt; CI job&lt;/strong&gt; (fresh venv, wheel install, real command run)&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;.github/workflows/ci-quality.yml&lt;/code&gt; (deleted per PROGRAM.md §2 / planning#57 — this repo runs no GitHub Actions) + &lt;a href=&quot;../../../tests/integration/test_clean_install_next.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/integration/test_clean_install_next.py&lt;/code&gt;&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Extend the &lt;strong&gt;same job&lt;/strong&gt; to install the published kernel/doctrine/charter wheels into the clean venv and exercise a charter command path. A new parallel job would fragment the gate.&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;WP12 supplies the groundwork this reuse rests on, and it is &lt;strong&gt;build-executed, not asserted&lt;/strong&gt;
(research.md §D7, spike status RESOLVED):&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/kernel/pyproject.toml&lt;/code&gt; — &lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt; 1.0.0, &lt;code dir=&quot;auto&quot;&gt;dependencies = []&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/pyproject.toml&lt;/code&gt; — declares &lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&gt;=1.0.0,&amp;#x3C;2.0.0&lt;/code&gt;, a real
import-closure need (&lt;code dir=&quot;auto&quot;&gt;doctrine/resolver.py&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;missions/primitives.py&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;shared/schema_utils.py&lt;/code&gt; all &lt;code dir=&quot;auto&quot;&gt;from kernel...&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/hatch_build.py&lt;/code&gt; — a hatchling
&lt;code dir=&quot;auto&quot;&gt;BuildHookInterface.initialize&lt;/code&gt; hook that force-includes the repo-root &lt;code dir=&quot;auto&quot;&gt;packs/&lt;/code&gt; tree as a
&lt;strong&gt;wheel-root sibling&lt;/strong&gt; of &lt;code dir=&quot;auto&quot;&gt;doctrine/&lt;/code&gt;, matching what &lt;code dir=&quot;auto&quot;&gt;pack_paths._resolve_built_in&lt;/code&gt;
expects (&lt;code dir=&quot;auto&quot;&gt;files(&quot;doctrine&quot;).parent / &quot;packs&quot; / &quot;built-in&quot;&lt;/code&gt;). A naive
&lt;code dir=&quot;auto&quot;&gt;force-include ../../packs&lt;/code&gt; escapes the project root and hatchling refuses it; a blanket
&lt;code dir=&quot;auto&quot;&gt;sources&lt;/code&gt; rename re-nests the hook’s own &lt;code dir=&quot;auto&quot;&gt;force_include&lt;/code&gt; entries under &lt;code dir=&quot;auto&quot;&gt;doctrine/&lt;/code&gt;. Both
were found by running the build, not by reading it.&lt;/li&gt;
&lt;li&gt;Proof: &lt;code dir=&quot;auto&quot;&gt;hatch build -t wheel&lt;/code&gt; in both trees, &lt;code dir=&quot;auto&quot;&gt;unzip -l&lt;/code&gt; confirming &lt;code dir=&quot;auto&quot;&gt;doctrine/&lt;/code&gt; and
&lt;code dir=&quot;auto&quot;&gt;packs/&lt;/code&gt; as true siblings, and an install-and-import closure run in a scratch venv where
&lt;code dir=&quot;auto&quot;&gt;resolve_pack_root(&quot;built-in&quot;)&lt;/code&gt; resolved against the installed sibling.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;d-deferred-follow-on-issue-set&quot;&gt;(d) Deferred follow-on issue set&lt;/h3&gt;
&lt;p&gt;The cutover follow-on owns this set. They are deferred, not dropped:&lt;/p&gt;








































&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Issue&lt;/th&gt;&lt;th&gt;Title&lt;/th&gt;&lt;th&gt;Relation to the cutover&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3101&quot;&gt;#3101&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Split &lt;code dir=&quot;auto&quot;&gt;src/doctrine/&lt;/code&gt; (and assess &lt;code dir=&quot;auto&quot;&gt;src/charter/&lt;/code&gt;) into a separate installable wheel/package&lt;/td&gt;&lt;td&gt;The cutover itself — the parent. This ADR is its assessment half.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3091&quot;&gt;#3091&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Phase 1b: relocate the &lt;code dir=&quot;auto&quot;&gt;missions/&lt;/code&gt; doctrine tree to &lt;code dir=&quot;auto&quot;&gt;packs/built-in&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Changes what the doctrine wheel must carry; sequence before or with the cutover.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3022&quot;&gt;#3022&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Extract built-in doctrine packs into &lt;code dir=&quot;auto&quot;&gt;spec-kitty-packs-open&lt;/code&gt;&lt;/td&gt;&lt;td&gt;A further split of the payload WP12’s build hook currently ships inside the doctrine wheel.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3036&quot;&gt;#3036&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Architectural gate requires the repo-coupling the shippable-doctrine rule forbids&lt;/td&gt;&lt;td&gt;A live contradiction between two gates; must be resolved before doctrine ships standalone.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/3039&quot;&gt;#3039&lt;/a&gt;&lt;/td&gt;&lt;td&gt;Split &lt;code dir=&quot;auto&quot;&gt;test_no_dead_doctrine_paths.py&lt;/code&gt;: Gate A is a &lt;code dir=&quot;auto&quot;&gt;src/&lt;/code&gt;-wide CLI gate wearing a doctrine name&lt;/td&gt;&lt;td&gt;Mis-scoped gate; would follow the wrong package after the split.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2986&quot;&gt;#2986&lt;/a&gt;&lt;/td&gt;&lt;td&gt;runtime→doctrine boundary ratchet sees only module-level imports; 61 function-local imports across 30 files bypass it&lt;/td&gt;&lt;td&gt;&lt;strong&gt;The same blind spot WP10 closed for charter&lt;/strong&gt;, still open for runtime→doctrine. Directly relevant: the boundary-test half of the extended pattern is unsound for that pair until it is fixed.&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;h3 id=&quot;e-c-002--this-mission-performed-no-wheel-cutover&quot;&gt;(e) C-002 — this Mission performed NO wheel cutover&lt;/h3&gt;
&lt;p&gt;Stated explicitly, because “packaging groundwork landed” is easy to misread as “the split
happened”:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The Mission &lt;code dir=&quot;auto&quot;&gt;doctrine-charter-split-unification-01KZ0SRB&lt;/code&gt; performed no wheel cutover. It
is groundwork only.&lt;/strong&gt; Concretely, as of this ADR:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/kernel&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;src/doctrine&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;src/charter&lt;/code&gt; &lt;strong&gt;remain in the root wheel&lt;/strong&gt; — the root
&lt;a href=&quot;../../../pyproject.toml&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;pyproject.toml&lt;/code&gt;&lt;/a&gt; &lt;code dir=&quot;auto&quot;&gt;packages&lt;/code&gt; list is untouched and still names
all three, and &lt;code dir=&quot;auto&quot;&gt;packs/&lt;/code&gt; is still root-force-included.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No wheel was published.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;spec-kitty-kernel&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;spec-kitty-doctrine&lt;/code&gt; exist as
manifests and build locally; they are not on PyPI.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No release gate was added, and no CI job builds or installs the nested wheels
standalone.&lt;/strong&gt; This is load-bearing: doctrine’s newly declared &lt;code dir=&quot;auto&quot;&gt;kernel&lt;/code&gt; dependency is
unresolvable until the follow-on publishes the kernel wheel, so a CI job building the
nested doctrine wheel today would break — which is why “groundwork only” is a CI-verified
constraint, not just an intention.&lt;/li&gt;
&lt;li&gt;There is deliberately &lt;strong&gt;no&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;src/charter/pyproject.toml&lt;/code&gt;. Minting one now would be the
first half of the forbidden partial cutover.&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;The follow-on starts from a settled assessment and a forced ordering, not a fresh
investigation; the remaining unknowns are named (glossary/runtime/re-export edges) rather
than assumed absent.&lt;/li&gt;
&lt;li&gt;The one real upward edge out of the charter layer is gone, and re-introducing it at any
scope now fails CI.&lt;/li&gt;
&lt;li&gt;The kernel/doctrine manifests are build-verified, so the follow-on inherits two solved
packaging problems (root-relative &lt;code dir=&quot;auto&quot;&gt;packages&lt;/code&gt;, out-of-tree &lt;code dir=&quot;auto&quot;&gt;packs/&lt;/code&gt;) instead of rediscovering
them.&lt;/li&gt;
&lt;li&gt;Enforcement stays in one idiom: three mechanisms, extended, not duplicated.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;The charter wheel is not available until the whole chain ships; consumers wanting only the
charter layer keep installing &lt;code dir=&quot;auto&quot;&gt;spec-kitty-cli&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;The no-partial rule makes the follow-on a large single reviewed unit, which is harder to
review than three small ones. This cost is accepted: #779 is the evidence that the cheaper
path is the expensive one.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/kernel/pyproject.toml&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;src/doctrine/pyproject.toml&lt;/code&gt; are dormant manifests in the
tree until the cutover — a standing “is this live?” question for readers, mitigated by this
ADR and by the closure test.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;The three-wheel end state is unchanged from #3101’s original intent; this ADR fixes the
order and the atomicity, not the destination.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;confirmation&quot;&gt;Confirmation&lt;/h3&gt;
&lt;p&gt;This decision is confirmed correct when the follow-on can execute mechanically: the extended
pyproject-shape assertions and the extended &lt;code dir=&quot;auto&quot;&gt;clean-install-verification&lt;/code&gt; job go green on a
fresh venv that installs the three published wheels and runs a charter command path, with no
new enforcement idiom introduced. It is falsified if the charter layer turns out to carry
non-import coupling to &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; that the AST gate never covered — which is exactly why
the claim above is scoped to import edges and why the unproven-edge table is part of the
decision rather than a footnote.&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-a--extract-the-charter-wheel-now&quot;&gt;Option A — Extract the charter wheel now&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Delivers a visible piece of #3101 immediately.&lt;/li&gt;
&lt;li&gt;Exercises the packaging path while WP10/WP12 context is fresh.&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;Unresolvable: charter’s 109 &lt;code dir=&quot;auto&quot;&gt;doctrine&lt;/code&gt; and 8 &lt;code dir=&quot;auto&quot;&gt;kernel&lt;/code&gt; import statements need both wheels
published first. Neither is.&lt;/li&gt;
&lt;li&gt;It is the #779 hybrid by construction — charter external, its dependencies still bundled
in the root wheel.&lt;/li&gt;
&lt;li&gt;Violates C-002 (no cutover this Mission) and the C-007 no-partial rule of
&lt;code dir=&quot;auto&quot;&gt;2026-04-25-1&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-b--assess-and-sequence-one-atomic-cutover-chosen&quot;&gt;Option B — Assess and sequence one atomic cutover (chosen)&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Matches the real dependency closure; the publish order is derived, not chosen.&lt;/li&gt;
&lt;li&gt;Reuses a landed, proven enforcement triad instead of inventing one.&lt;/li&gt;
&lt;li&gt;Keeps the groundwork honest: the manifests exist and build, and nothing pretends the split
has happened.&lt;/li&gt;
&lt;li&gt;Names the unproven couplings so the follow-on scopes its own assessment correctly.&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;Defers all user-visible packaging value.&lt;/li&gt;
&lt;li&gt;Concentrates risk in one large follow-on PR.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-c--abandon-the-split&quot;&gt;Option C — Abandon the split&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Zero further cost; the monorepo wheel works today.&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;Forfeits the independent doctrine/charter distribution that #3022 and #3091 depend on.&lt;/li&gt;
&lt;li&gt;Wastes the landed WP10/WP12 groundwork.&lt;/li&gt;
&lt;li&gt;Leaves the layer chain a convention enforced only by tests, never by package boundaries.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Extends: ADR &lt;a href=&quot;/adr/3.x/2026-04-25-1-shared-package-boundary/&quot;&gt;2026-04-25-1 — Shared Package Boundary Cutover&lt;/a&gt;
(the boundary-test + pyproject-shape-test + &lt;code dir=&quot;auto&quot;&gt;clean-install-verification&lt;/code&gt; pattern, and the
C-007 no-partial rule)&lt;/li&gt;
&lt;li&gt;Mission spec: &lt;code dir=&quot;auto&quot;&gt;kitty-specs/doctrine-charter-split-unification-01KZ0SRB/spec.md&lt;/code&gt;
(FR-011, SC-007, C-002, User Story 4)&lt;/li&gt;
&lt;li&gt;Research: &lt;code dir=&quot;auto&quot;&gt;kitty-specs/doctrine-charter-split-unification-01KZ0SRB/research.md&lt;/code&gt;
(§D7 executed &lt;code dir=&quot;auto&quot;&gt;hatch build&lt;/code&gt; spike result, §D8 sequencing DAG, §D10 claim-scope correction)&lt;/li&gt;
&lt;li&gt;Landed enforcement and groundwork:
&lt;ul&gt;
&lt;li&gt;&lt;a href=&quot;../../../tests/architectural/test_charter_no_specify_cli_import.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_charter_no_specify_cli_import.py&lt;/code&gt;&lt;/a&gt; — the AST-walk gate this ADR’s claim rests on&lt;/li&gt;
&lt;li&gt;&lt;a href=&quot;../../../tests/architectural/test_layer_rules.py&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_layer_rules.py&lt;/code&gt;&lt;/a&gt; — the pytestarch layer rule; does not catch imports nested inside &lt;code dir=&quot;auto&quot;&gt;try&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;except&lt;/code&gt; handlers (the exact shape of the deleted edge), so it stayed green with the violation present&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_doctrine_wheel_closure.py&lt;/code&gt; — non-vacuous doctrine manifest closure (removed along with the rest of this dormant wheel groundwork by mission &lt;code dir=&quot;auto&quot;&gt;charter-code-topology-01M152G1&lt;/code&gt;)&lt;/li&gt;
&lt;li&gt;&lt;a href=&quot;../../../src/kernel/pyproject.toml&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;src/kernel/pyproject.toml&lt;/code&gt;&lt;/a&gt;, &lt;code dir=&quot;auto&quot;&gt;src/doctrine/pyproject.toml&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;src/doctrine/hatch_build.py&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;Runbook precedent: &lt;a href=&quot;/migrations/shared-package-boundary-cutover/&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;docs/migrations/shared-package-boundary-cutover.md&lt;/code&gt;&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
