<?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/2026-06-11-1-op-as-first-class-execution-artifact/">
  <id>https://docs.spec-kitty.ai/adr/3.x/2026-06-11-1-op-as-first-class-execution-artifact/</id>
  <title>ADR 2026-06-11-1: Op as a First-Class Execution Artifact (Mission ⟷ Op ⟷ ad-hoc)</title>
  <updated>2026-06-11T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/3.x/2026-06-11-1-op-as-first-class-execution-artifact/"/>
  <summary>Names the Op — a bounded, doctrine-governed, immediately dispatched action with a durable record — as the one primitive shared by ad-hoc dispatch and lifecycle flows.</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;code dir=&quot;auto&quot;&gt;spec-kitty dispatch&lt;/code&gt; opens bounded, doctrine-governed agent actions that run
immediately — no spec, no plan, no work
packages, no lane worktree — yet they still load governance context, route to an
agent profile, and produce a durable record. The implementation lives in
&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/invocation/&lt;/code&gt; and already emits &lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt; /
&lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt; (see &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/invocation/__init__.py&lt;/code&gt;), but the
&lt;strong&gt;concept&lt;/strong&gt; these commands express has never been named or ratified at the
architecture level. #1688 (the proposal, now closed/superseded into this ADR)
diagnosed the gap precisely: the records were treated as scratch byproducts of
“governance machinery” rather than as instances of a real unit of work.&lt;/p&gt;
&lt;p&gt;Two epics now press on the same gap from different directions:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;#1804 (Ops execution layer)&lt;/strong&gt; wants a durable, governed, queue-backed
execution artifact for standalone dispatch. It has &lt;strong&gt;no ADR&lt;/strong&gt; — it is the highest-value
architecture gap in &lt;code dir=&quot;auto&quot;&gt;work/EPIC_ARCHITECTURE_CORRELATION.md&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;#1802 (pre/post-mission lifecycle)&lt;/strong&gt; wants bounded, governed flows that sit
&lt;em&gt;outside&lt;/em&gt; a Mission’s spec→merge loop: an &lt;strong&gt;intake&lt;/strong&gt; flow before a Mission
exists and a &lt;strong&gt;correction&lt;/strong&gt; flow after a Mission has merged.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Both want the same thing: &lt;em&gt;a bounded, doctrine-governed action with a durable
record and no full Mission lifecycle.&lt;/em&gt; Left unguided, #1804 and #1802 would each
mint their own primitive — two parallel abstractions for one shape. The
mission’s binding constraint &lt;strong&gt;C-005 (single source of truth — no parallel
narrative/abstraction surfaces)&lt;/strong&gt; forbids that. This ADR is the seam that
prevents the divergence: it names the shared primitive &lt;strong&gt;once&lt;/strong&gt; so both epics
build on it rather than beside it.&lt;/p&gt;
&lt;p&gt;The decision matters now because &lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt; are already
in the codebase. Ratifying the concept before #1804/#1802 land code is the
cheapest moment to converge; after, we would be unifying two shipped primitives.&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;C-005 — one abstraction, not two.&lt;/strong&gt; #1804 (Ops) and #1802 (lifecycle) must
share a single execution primitive.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Screaming the tier.&lt;/strong&gt; The tier between a full Mission and an ungoverned
ad-hoc shell command must be a named, discoverable concept, not an emergent
property of three sibling commands.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Durability + governance without the full planning loop.&lt;/strong&gt; The primitive must carry real
doctrine context and a permanent trace, while staying lighter than a Mission
(no spec/plan/tasks/worktree/merge).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Coherence with the current execution shapes.&lt;/strong&gt; Any execution-context,
commit, or status mechanics the Op touches must reference the &lt;em&gt;canonical&lt;/em&gt;
surfaces (&lt;code dir=&quot;auto&quot;&gt;mission_runtime&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;core/commit_guard&lt;/code&gt;, the append-only status
event log), never the retired ones.&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 — Two primitives (status quo drift):&lt;/strong&gt; let #1804 build an “Ops”
artifact and #1802 build a separate “lifecycle flow” artifact.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option B — One shared primitive: the Op (chosen).&lt;/strong&gt; Ratify the Op as the
single tier between Mission and ad-hoc; pre/post-mission lifecycle flows are
Ops, not a parallel concept.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option C — Fold everything into Mission:&lt;/strong&gt; model intake/correction/dispatch
as degenerate Missions.&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 — One shared primitive: the Op.”&lt;/strong&gt;&lt;/p&gt;
&lt;h3 id=&quot;what-an-op-is&quot;&gt;What an Op is&lt;/h3&gt;
&lt;p&gt;An &lt;strong&gt;Op&lt;/strong&gt; is a &lt;em&gt;bounded, doctrine-governed agent action, dispatched immediately,
that produces a durable governed record without a full Mission lifecycle.&lt;/em&gt; It is
to &lt;code dir=&quot;auto&quot;&gt;dispatch&lt;/code&gt; what a Mission is to the
spec→plan→tasks→implement→review→merge loop: lighter, immediate, no planning
loop — but real work, real doctrine, real record.&lt;/p&gt;
&lt;p&gt;The Op lifecycle is a short dispatch loop, not a Mission state machine:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Route&lt;/strong&gt; — resolve the operator request to an agent profile (explicit
&lt;code dir=&quot;auto&quot;&gt;--profile&lt;/code&gt;, else the router). Ambiguous handles fail closed (no silent
fallback), consistent with the identity-selector rule.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Load governance context&lt;/strong&gt; — assemble the action-scoped doctrine payload for
the resolved profile (the same charter-context surface a Mission step uses).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Open a durable Op record&lt;/strong&gt; — append an &lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt; to the append-only
audit trail (&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/invocation/&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Agent acts on the panel&lt;/strong&gt; — the dispatched agent does the work under the
loaded governance context; the CLI does not perform the work itself.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Close the Op&lt;/strong&gt; — append an &lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt; with the real outcome.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;An Op carries a &lt;code dir=&quot;auto&quot;&gt;ModeOfWork&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;task_execution&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;mission_step&lt;/code&gt;, or &lt;code dir=&quot;auto&quot;&gt;query&lt;/code&gt;, per
&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/invocation/modes.py&lt;/code&gt;). The Op record is the durable trace; durability is a
&lt;em&gt;property of the artifact&lt;/em&gt;, not a bolt-on. (#1688 framed the original
two-invocation durability bug as a symptom of this missing concept; naming the Op
resolves it structurally.)&lt;/p&gt;
&lt;h3 id=&quot;the-three-execution-tiers&quot;&gt;The three execution tiers&lt;/h3&gt;

































&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Tier&lt;/th&gt;&lt;th&gt;Lifecycle&lt;/th&gt;&lt;th&gt;Governance&lt;/th&gt;&lt;th&gt;Durable record&lt;/th&gt;&lt;th&gt;Workspace&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;Mission&lt;/strong&gt;&lt;/td&gt;&lt;td&gt;spec → plan → tasks → implement → review → merge&lt;/td&gt;&lt;td&gt;full doctrine + step contracts&lt;/td&gt;&lt;td&gt;mission artifacts + status event log&lt;/td&gt;&lt;td&gt;lane worktrees + mission branch&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;Op&lt;/strong&gt;&lt;/td&gt;&lt;td&gt;route → context → record → act → close&lt;/td&gt;&lt;td&gt;action-scoped doctrine payload&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt; trail&lt;/td&gt;&lt;td&gt;none (acts in the current checkout)&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;strong&gt;ad-hoc&lt;/strong&gt;&lt;/td&gt;&lt;td&gt;none&lt;/td&gt;&lt;td&gt;none&lt;/td&gt;&lt;td&gt;none&lt;/td&gt;&lt;td&gt;none&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;The Op is the &lt;strong&gt;governed middle tier&lt;/strong&gt;. The distinction from ad-hoc is exactly
governance + durability: an ad-hoc shell command is ungoverned and untraced; an
Op is governed and traced. The distinction from a Mission is the absence of the
planning loop, work packages, and dedicated worktrees.&lt;/p&gt;
&lt;h3 id=&quot;prepost-mission-lifecycle-flows-are-ops-the-c-005-unification&quot;&gt;Pre/post-mission lifecycle flows are Ops (the C-005 unification)&lt;/h3&gt;
&lt;p&gt;The binding clause of this ADR: &lt;strong&gt;pre/post-mission lifecycle actions (#1802) are
Ops, not a second primitive.&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Intake Op&lt;/strong&gt; — a pre-mission flow (e.g. triaging a request into a future
Mission) is an Op: bounded, governed, durable-recorded, no Mission yet.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Correction Op&lt;/strong&gt; — a post-merge flow (e.g. a governed follow-up fix after a
Mission has merged) is an Op: bounded, governed, durable-recorded, no new
Mission.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Epics #1804 (Ops) and #1802 (lifecycle) therefore consume &lt;strong&gt;one&lt;/strong&gt; abstraction.
Epic #1804 builds the Op artifact and its queue-backed durability; #1802’s
intake/correction flows are &lt;em&gt;modes of dispatching an Op&lt;/em&gt;, not a parallel artifact
type. Neither epic may mint a competing primitive — doing so reintroduces the
C-005 violation this ADR exists to prevent.&lt;/p&gt;
&lt;h3 id=&quot;coherence-with-the-canonical-execution-shapes-binding&quot;&gt;Coherence with the canonical execution shapes (binding)&lt;/h3&gt;
&lt;p&gt;Where an Op touches execution mechanics, it consumes the &lt;strong&gt;current canonical
surfaces&lt;/strong&gt; — never the retired ones:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Execution context.&lt;/strong&gt; If an Op needs resolved context, it calls
&lt;code dir=&quot;auto&quot;&gt;resolve_action_context&lt;/code&gt; from &lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;mission_runtime&lt;/code&gt;&lt;/strong&gt;
(&lt;code dir=&quot;auto&quot;&gt;src/mission_runtime/resolution.py&lt;/code&gt;), which returns the immutable
&lt;code dir=&quot;auto&quot;&gt;ExecutionContext&lt;/code&gt; value object (&lt;code dir=&quot;auto&quot;&gt;src/mission_runtime/context.py&lt;/code&gt;). The retired
&lt;code dir=&quot;auto&quot;&gt;specify_cli/core/execution_context.py&lt;/code&gt; path and the value-object’s old
home are historical — see ADR
&lt;a href=&quot;/adr/3.x/2026-06-07-1-execution-state-canonical-surface/&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;2026-06-07-1-execution-state-canonical-surface.md&lt;/code&gt;&lt;/a&gt;
and the 2026-06-10 addendum to
&lt;a href=&quot;/adr/3.x/2026-06-03-2-executioncontext-owner-and-committarget/&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;2026-06-03-2-executioncontext-owner-and-committarget.md&lt;/code&gt;&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Commit safety.&lt;/strong&gt; Ops run in the operator’s current checkout, so most Ops do
not commit at all. Any Op that &lt;em&gt;does&lt;/em&gt; create a commit routes its protection
decision through the single policy seam &lt;code dir=&quot;auto&quot;&gt;core.commit_guard.evaluate&lt;/code&gt;
(&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/core/commit_guard.py&lt;/code&gt;) — the &lt;strong&gt;one&lt;/strong&gt; protected-branch
decision (C-GUARD-1). The destination is carried as the canonical
&lt;code dir=&quot;auto&quot;&gt;CommitTarget(ref: str, kind: CommitTargetKind)&lt;/code&gt; value object
(&lt;code dir=&quot;auto&quot;&gt;kind ∈ {PRIMARY, COORDINATION, FLATTENED}&lt;/code&gt;), and authorization is the
asserted-at-the-surface &lt;code dir=&quot;auto&quot;&gt;GuardCapability&lt;/code&gt; parameter — never derived from
message text, file content, env, or op records. An Op carries no special
protected-branch privilege: its default capability is
&lt;code dir=&quot;auto&quot;&gt;GuardCapability.STANDARD&lt;/code&gt;. This is the delivered &lt;code dir=&quot;auto&quot;&gt;(ref, kind)&lt;/code&gt; /
&lt;code dir=&quot;auto&quot;&gt;GuardCapability&lt;/code&gt; shape (Strangler step 7, delivered across missions 01KTPKST +
01KTRC04), &lt;strong&gt;not&lt;/strong&gt; the retired &lt;code dir=&quot;auto&quot;&gt;(worktree_root, destination_ref)&lt;/code&gt; sketch.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Status.&lt;/strong&gt; An Op never drives the 9-lane WP status machine; the append-only
Op event trail is its sole record. WP lane state remains the exclusive
authority of the status event log and is untouched by Ops.&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;#1804 and #1802 build on one ratified primitive — C-005 holds; no parallel
abstraction.&lt;/li&gt;
&lt;li&gt;The execution layer gains a named, discoverable middle tier; the “missing
concept” #1688 identified is closed.&lt;/li&gt;
&lt;li&gt;Op durability is structural (the artifact owns its trail), so the original
two-invocation durability failures cannot recur as one-off patches.&lt;/li&gt;
&lt;li&gt;The Op’s mechanics are pinned to the canonical execution surfaces, so the
layer cannot drift back onto the retired &lt;code dir=&quot;auto&quot;&gt;execution_context.py&lt;/code&gt; /
&lt;code dir=&quot;auto&quot;&gt;(worktree_root, destination_ref)&lt;/code&gt; shapes.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt; and the &lt;code dir=&quot;auto&quot;&gt;invocation/&lt;/code&gt; package predate this
ratification; #1804 must reconcile the shipped event/record shapes with the
ratified concept (rename/relocate work is implementation scope, tracked under
#1804, not this ADR).&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;This ADR ratifies the &lt;em&gt;concept and boundaries&lt;/em&gt;; the durable queue-backed
storage mechanics (#1804) and the intake/correction dispatch surfaces (#1802)
are designed in their own implementation specs, governed by this seam.&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 #1804 and #1802 both reference this ADR as their
architecture anchor and neither introduces a competing execution primitive — i.e.
&lt;code dir=&quot;auto&quot;&gt;work/EPIC_ARCHITECTURE_CORRELATION.md&lt;/code&gt; shows the #1804 Ops gap closed (mission
success criterion SC-2) and #1802 listed against this ADR. Confidence: high —
the Op shape is already partially expressed in shipped code (&lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt; /
&lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;ModeOfWork&lt;/code&gt;), so this ratifies an emergent reality rather
than proposing a speculative one.&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--two-primitives-status-quo-drift&quot;&gt;Option A — Two primitives (status-quo drift)&lt;/h3&gt;
&lt;p&gt;Let #1804 and #1802 each define their own artifact.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;No coordination needed between the two epics up front.&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;Direct C-005 violation: two abstractions for one shape.&lt;/li&gt;
&lt;li&gt;Guarantees future consolidation debt (unifying two shipped primitives).&lt;/li&gt;
&lt;li&gt;The execution-layer concept stays unnamed and undiscoverable.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-b--one-shared-primitive-the-op-chosen&quot;&gt;Option B — One shared primitive: the Op (chosen)&lt;/h3&gt;
&lt;p&gt;Ratify the Op as the single Mission ⟷ ad-hoc tier; lifecycle flows are Ops.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Satisfies C-005 — one primitive, two consumers.&lt;/li&gt;
&lt;li&gt;Names a real, already-emergent concept; closes the #1688 gap structurally.&lt;/li&gt;
&lt;li&gt;Pins Op mechanics to canonical execution surfaces.&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;Requires #1804 to reconcile pre-existing &lt;code dir=&quot;auto&quot;&gt;invocation/&lt;/code&gt; shapes with the
ratified concept.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;option-c--fold-everything-into-mission&quot;&gt;Option C — Fold everything into Mission&lt;/h3&gt;
&lt;p&gt;Model intake/correction/dispatch as degenerate Missions.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Reuses the existing Mission machinery.&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;Forces the full planning loop (spec/plan/tasks/worktree/merge) onto bounded
immediate actions — the exact weight the Op tier exists to avoid.&lt;/li&gt;
&lt;li&gt;Pollutes the Mission state machine and status authority with non-Mission work.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;adjudications-wp06--escalated-from-wp02s-approved-review&quot;&gt;Adjudications (WP06 — escalated from WP02’s approved review)&lt;/h2&gt;
&lt;p&gt;Two cross-cutting items in the architecture surface were escalated for the
architect to adjudicate. They are recorded here because they belong to the
&lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/**&lt;/code&gt; authority this WP owns.&lt;/p&gt;
&lt;h3 id=&quot;adjudication-1--charter-authority-path-modernization-deferred&quot;&gt;Adjudication 1 — Charter authority-path modernization: &lt;strong&gt;deferred&lt;/strong&gt;&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Finding.&lt;/strong&gt; The charter’s default authority pointer for architectural intent
(&lt;code dir=&quot;auto&quot;&gt;charter.context_renderers.authority_paths.DEFAULT_AUTHORITY_PATHS&lt;/code&gt;) still cites
&lt;code dir=&quot;auto&quot;&gt;architecture/2.x/adr/&lt;/code&gt;, while &lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/&lt;/code&gt; is the canonical track.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Decision: defer the flip, with a sanctioned recovery path.&lt;/strong&gt; Rationale:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Not broken today.&lt;/strong&gt; WP02’s living layout converted the post-cutover 2.x ADR
paths into back-compat &lt;strong&gt;symlinks into &lt;code dir=&quot;auto&quot;&gt;3.x/&lt;/code&gt;&lt;/strong&gt;, and the renderer is
existence-gated. So &lt;code dir=&quot;auto&quot;&gt;architecture/2.x/adr/&lt;/code&gt; still resolves; the pointer is a
back-compat &lt;em&gt;alias&lt;/em&gt;, not a dangling link. The modernization is a clarity
improvement, not a correctness fix.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The true blast radius exceeds the sanctioned chain.&lt;/strong&gt; The flip’s enumerated
chain was: &lt;code dir=&quot;auto&quot;&gt;authority_paths.py&lt;/code&gt; → the two software-dev SOURCE prompt templates
→ regenerated agent copies → two contract tests
(&lt;code dir=&quot;auto&quot;&gt;test_template_governance_payload_contract.py&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;test_wp_prompt_governance_contract.py&lt;/code&gt;) → the twelve-agent parity baselines.
Investigation found &lt;strong&gt;three additional surfaces outside that chain&lt;/strong&gt; that
assert the &lt;code dir=&quot;auto&quot;&gt;2.x/adr/&lt;/code&gt; default and would break on a flip:
&lt;code dir=&quot;auto&quot;&gt;tests/charter/test_context_authority_paths.py&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;tests/charter/test_sync_authority_paths.py&lt;/code&gt;, and
&lt;code dir=&quot;auto&quot;&gt;tests/charter/test_schemas_additive_fields.py&lt;/code&gt;. It also requires
re-coordinating &lt;code dir=&quot;auto&quot;&gt;.kittify/charter/charter.md&lt;/code&gt; — a file &lt;strong&gt;WP02 owns&lt;/strong&gt; and which
WP02 deliberately left annotated as &lt;code dir=&quot;auto&quot;&gt;architecture/2.x/adr/  # 2.x-era architectural decisions (historical)&lt;/code&gt; alongside a separate active
&lt;code dir=&quot;auto&quot;&gt;architecture/adrs/&lt;/code&gt; entry. Flipping the default unilaterally from WP06 would
reach across the WP02 ownership boundary and beyond the enumerated chain.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Per the WP06 mandate&lt;/strong&gt;, when the flip requires surfaces beyond the
enumerated chain, the sanctioned action is to record a reasoned deferral here
rather than improvise a partial, test-breaking flip.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;strong&gt;Recovery path (when undertaken).&lt;/strong&gt; Do all links together, in one change:
flip &lt;code dir=&quot;auto&quot;&gt;DEFAULT_AUTHORITY_PATHS&lt;/code&gt; and its docstrings → update both software-dev
SOURCE prompt templates (&lt;code dir=&quot;auto&quot;&gt;implement/prompt.md&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;review/prompt.md&lt;/code&gt;) → regenerate
agent copies via the documented upgrade flow → update the two governance-contract
tests → update the three &lt;code dir=&quot;auto&quot;&gt;tests/charter/&lt;/code&gt; assertions → re-coordinate
&lt;code dir=&quot;auto&quot;&gt;.kittify/charter/charter.md&lt;/code&gt; with the owning WP → regenerate the twelve-agent
parity baselines (&lt;code dir=&quot;auto&quot;&gt;PYTEST_UPDATE_SNAPSHOTS=1 pytest tests/specify_cli/regression/&lt;/code&gt;)
and commit the baselines &lt;em&gt;with&lt;/em&gt; the template change → confirm
&lt;code dir=&quot;auto&quot;&gt;pytest tests/architectural/&lt;/code&gt; green. This is a coherent follow-up Op/mission, not
a WP06 side effect.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Addendum — 2026-06-12 — EXECUTED&lt;/strong&gt; (WP07, mission &lt;code dir=&quot;auto&quot;&gt;name-vs-authority-remediation-01KTYGTE&lt;/code&gt;):
The deferred flip was executed in full. All seven chain links landed atomically:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/charter/context_renderers/authority_paths.py&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;DEFAULT_AUTHORITY_PATHS&lt;/code&gt; flipped
&lt;code dir=&quot;auto&quot;&gt;architecture/2.x/adr/&lt;/code&gt; → &lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/&lt;/code&gt;; module docstring and dict docstring updated.&lt;/li&gt;
&lt;li&gt;Both source prompts updated: &lt;code dir=&quot;auto&quot;&gt;src/doctrine/missions/mission-steps/software-dev/implement/prompt.md&lt;/code&gt;
and &lt;code dir=&quot;auto&quot;&gt;review/prompt.md&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Two governance-contract tests updated: &lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_template_governance_payload_contract.py&lt;/code&gt;
(fixture directory + two path assertions) and &lt;code dir=&quot;auto&quot;&gt;tests/specify_cli/next/test_wp_prompt_governance_contract.py&lt;/code&gt;
(inline charter string, &lt;code dir=&quot;auto&quot;&gt;adr_path_present&lt;/code&gt; check, self-sufficiency regex extended to &lt;code dir=&quot;auto&quot;&gt;[23]\.x&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;Three &lt;code dir=&quot;auto&quot;&gt;tests/charter/&lt;/code&gt; assertions updated: &lt;code dir=&quot;auto&quot;&gt;test_context_authority_paths.py&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;test_sync_authority_paths.py&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;test_schemas_additive_fields.py&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;.kittify/charter/charter.md&lt;/code&gt; line 317 annotation updated
(&lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/ # canonical architectural decisions (3.x era)&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;Twelve-agent parity baselines regenerated via
&lt;code dir=&quot;auto&quot;&gt;PYTEST_UPDATE_SNAPSHOTS=1 pytest tests/specify_cli/regression/&lt;/code&gt; — 26 baseline files
updated (13 agents × implement + review); no other baseline churn observed.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;pytest tests/architectural/ -q&lt;/code&gt; fully green post-flip.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;No 2.x-pointing authority default remains in any active code path.&lt;/p&gt;
&lt;h3 id=&quot;adjudication-2--duplicate-doctrine-layer-merge-semantics-adr-3x-is-canonical-already-resolved-by-wp02&quot;&gt;Adjudication 2 — Duplicate &lt;code dir=&quot;auto&quot;&gt;doctrine-layer-merge-semantics&lt;/code&gt; ADR: &lt;strong&gt;3.x is canonical (already resolved by WP02)&lt;/strong&gt;&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Finding.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;2026-05-16-1-doctrine-layer-merge-semantics.md&lt;/code&gt; existed in both
&lt;code dir=&quot;auto&quot;&gt;2.x/adr/&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;3.x/adr/&lt;/code&gt;; &lt;code dir=&quot;auto&quot;&gt;docs/explanation/org-doctrine-layer.md&lt;/code&gt; linked the
2.x copy.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Decision: &lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/&lt;/code&gt; is canonical for this ADR; the 2.x path is a
back-compat pointer.&lt;/strong&gt; This is consistent with the era rule WP02 recorded in
&lt;code dir=&quot;auto&quot;&gt;architecture/3.x/adr/README.md&lt;/code&gt; (“This folder is canonical for 3.x decisions.
Back-compat symlinks at the old &lt;code dir=&quot;auto&quot;&gt;architecture/2.x/adr/&amp;#x3C;filename&gt;&lt;/code&gt; paths point
here”). WP02 already converted the 2.x copy into a &lt;strong&gt;symlink&lt;/strong&gt; into &lt;code dir=&quot;auto&quot;&gt;3.x/&lt;/code&gt;, so
the duplication is structurally resolved — there is one canonical file. The only
residual action is editorial: the stale 2.x link in
&lt;code dir=&quot;auto&quot;&gt;docs/explanation/org-doctrine-layer.md&lt;/code&gt; is repointed at the canonical 3.x path
(done as part of this WP, with a one-line out-of-map rationale).&lt;/p&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Mission spec (FR-007): &lt;code dir=&quot;auto&quot;&gt;kitty-specs/doctrine-glossary-architecture-consolidation-01KTNWFC/spec.md&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;Proposal (superseded into this ADR): &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1688&quot;&gt;#1688&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Consuming epics: &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1804&quot;&gt;#1804&lt;/a&gt; (Ops), &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1802&quot;&gt;#1802&lt;/a&gt; (lifecycle), &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/1810&quot;&gt;#1810&lt;/a&gt; (dispatch collapse)&lt;/li&gt;
&lt;li&gt;Current Op surface in code: &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/invocation/&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;OpStartedEvent&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;OpCompletedEvent&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;ModeOfWork&lt;/code&gt;), &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/cli/commands/{dispatch,profile_invocation,invocations_cmd}.py&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;Canonical execution surfaces: ADR &lt;a href=&quot;/adr/3.x/2026-06-07-1-execution-state-canonical-surface/&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;2026-06-07-1-execution-state-canonical-surface.md&lt;/code&gt;&lt;/a&gt; (&lt;code dir=&quot;auto&quot;&gt;mission_runtime&lt;/code&gt;); ADR &lt;a href=&quot;/adr/3.x/2026-06-03-2-executioncontext-owner-and-committarget/&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;2026-06-03-2-executioncontext-owner-and-committarget.md&lt;/code&gt;&lt;/a&gt; + its 2026-06-10 addendum (&lt;code dir=&quot;auto&quot;&gt;CommitTarget(ref, kind)&lt;/code&gt;, step 7 delivered); &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/core/commit_guard.py&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;GuardCapability&lt;/code&gt;, single &lt;code dir=&quot;auto&quot;&gt;evaluate&lt;/code&gt;)&lt;/li&gt;
&lt;li&gt;Correlation matrix: &lt;code dir=&quot;auto&quot;&gt;work/EPIC_ARCHITECTURE_CORRELATION.md&lt;/code&gt; (Ops gap → SC-2)&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
