<?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/2026-07-16-2-mission-type-step-authority-and-template-vocabulary/">
  <id>https://docs.spec-kitty.ai/adr/3.x/2026-07-16-2-mission-type-step-authority-and-template-vocabulary/</id>
  <title>ADR: Steps are the mission-type building block; &quot;template&quot; is a doctrine artefact kind</title>
  <updated>2026-07-16T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/3.x/2026-07-16-2-mission-type-step-authority-and-template-vocabulary/"/>
  <summary>A step is the mission-type building block — the existing ACTION node enriched, not a new node kind — while `template` narrows to one doctrine artefact kind.</summary>
  <content type="html">&lt;h2 id=&quot;steps-are-the-mission-type-building-block-template-is-a-doctrine-artefact-kind&quot;&gt;Steps are the mission-type building block; “template” is a doctrine artefact kind&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Status:&lt;/strong&gt; Proposed · &lt;strong&gt;Epic:&lt;/strong&gt; #2652 (mission-type unification) · &lt;strong&gt;Builds on:&lt;/strong&gt; ADR
&lt;code dir=&quot;auto&quot;&gt;2026-07-15-1-doctrine-offers-charter-activates-runtime-consumes&lt;/code&gt; (S0/S4), the merged #2712 (mission_type as a
DRG node + mission_type→action edges) and #2689 (&lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; slot). &lt;strong&gt;Grounding:&lt;/strong&gt; design + code +
adversarial squad (see &lt;code dir=&quot;auto&quot;&gt;docs/plans/investigations/mission-type-step-model-unification.md&lt;/code&gt;).&lt;/p&gt;
&lt;h2 id=&quot;context&quot;&gt;Context&lt;/h2&gt;
&lt;p&gt;Three forces converged into a naming-and-authority problem:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;“template” is an overloaded discriminator.&lt;/strong&gt; The word names three disjoint things: &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt;
(a per-type flat map of &lt;code dir=&quot;auto&quot;&gt;spec&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;plan&lt;/code&gt; → &lt;em&gt;scaffold filenames&lt;/em&gt;, #2689), the &lt;code dir=&quot;auto&quot;&gt;template:&lt;/code&gt; DRG artefact kind
(16 reusable &lt;em&gt;doctrine exemplars&lt;/em&gt;, #2712), and per-step &lt;em&gt;prompt/WP templates&lt;/em&gt;. Using “template” as the thing
a mission type is &lt;em&gt;made of&lt;/em&gt; conflates a content scaffold with mission-type structure and collides nominally
with the doctrine artefact kind.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;A step split-brain.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;mission_types/&amp;#x3C;type&gt;.yaml&lt;/code&gt; authors a &lt;strong&gt;flat&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt; + &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt;,
while the rich per-step data already lives in a &lt;strong&gt;separate&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;mission-steps/&amp;#x3C;type&gt;/&amp;#x3C;step&gt;/step.yaml&lt;/code&gt; tree
parsed into the live &lt;code dir=&quot;auto&quot;&gt;MissionStep&lt;/code&gt; model (&lt;code dir=&quot;auto&quot;&gt;doctrine/missions/models.py:87-125&lt;/code&gt;, carrying &lt;code dir=&quot;auto&quot;&gt;prompt_template&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;agent_profile&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;depends_on&lt;/code&gt;) and consumed by the runtime. The DRG extractor reads only the
flat &lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;extractor.py:835&lt;/code&gt;) and never sees &lt;code dir=&quot;auto&quot;&gt;step.yaml&lt;/code&gt;. Two authoring surfaces, one
structure — the step model is ~70% shipped but not the authority.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;A fail-closed regression.&lt;/strong&gt; After #2689, &lt;code dir=&quot;auto&quot;&gt;documentation&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;research&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;plan&lt;/code&gt; (all default-activated) carry
&lt;code dir=&quot;auto&quot;&gt;template_set: null&lt;/code&gt;, so creating a mission of those types hard-fails (&lt;code dir=&quot;auto&quot;&gt;resolver.py:395&lt;/code&gt;). Spec-deliberate,
but three of four built-in types are uncreatable until their content templates are authored.&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;decision&quot;&gt;Decision&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Adopt a single vocabulary and a single authority.&lt;/strong&gt;&lt;/p&gt;
&lt;h3 id=&quot;d1--template-is-a-doctrine-artefact-kind-narrowed-canonical&quot;&gt;D1 — “template” is a doctrine artefact kind (narrowed, canonical)&lt;/h3&gt;
&lt;p&gt;A &lt;strong&gt;template&lt;/strong&gt; is a doctrine artefact that is an &lt;em&gt;example/skeleton&lt;/em&gt; — the existing &lt;code dir=&quot;auto&quot;&gt;NodeKind.TEMPLATE&lt;/code&gt;
population, one artefact kind among directive/tactic/procedure/paradigm/styleguide/toolguide/asset. “Template”
stops being a mission-type structural term.&lt;/p&gt;
&lt;h3 id=&quot;d2--a-step-is-the-mission-type-building-block-and-a-step-is-the-existing-actionmissionstep&quot;&gt;D2 — a &lt;strong&gt;step&lt;/strong&gt; is the mission-type building block, and a step **is the existing &lt;code dir=&quot;auto&quot;&gt;ACTION&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;MissionStep&lt;/code&gt;&lt;/h3&gt;
&lt;p&gt;enriched — NOT a new node kind.** Action nodes already &lt;em&gt;are&lt;/em&gt; the steps (one per &lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt; entry, with
&lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt; in-edges from the mission type and &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt; out-edges to their governing doctrine). A mission type is
an ordered set of steps; each step carries: &lt;code dir=&quot;auto&quot;&gt;description&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;recommended_role?&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;recommended_model_tier?&lt;/code&gt;, a
parameterizable &lt;code dir=&quot;auto&quot;&gt;prompt&lt;/code&gt;, an optional &lt;code dir=&quot;auto&quot;&gt;template&lt;/code&gt; reference, and (later) &lt;code dir=&quot;auto&quot;&gt;substeps&lt;/code&gt;.&lt;/p&gt;
&lt;h3 id=&quot;d3--one-template-kind-the-relation-encodes-the-lifecycle&quot;&gt;D3 — one &lt;code dir=&quot;auto&quot;&gt;template:&lt;/code&gt; kind; the &lt;strong&gt;relation&lt;/strong&gt; encodes the lifecycle&lt;/h3&gt;
&lt;p&gt;A step &lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;instantiates&lt;/code&gt;&lt;/strong&gt; a fill-once &lt;em&gt;skeleton&lt;/em&gt; template (&lt;code dir=&quot;auto&quot;&gt;Relation.INSTANTIATES&lt;/code&gt;, already exists); tactics and
directives &lt;strong&gt;reference&lt;/strong&gt; a read-only &lt;em&gt;exemplar&lt;/em&gt; template. Same &lt;code dir=&quot;auto&quot;&gt;NodeKind.TEMPLATE&lt;/code&gt;, different relation — so
there is no “fourth template,” and the two lifecycles (mission-emitted output vs referenced doctrine exemplar)
stay distinguishable. This finishes the &lt;code dir=&quot;auto&quot;&gt;mission_type → step → template&lt;/code&gt; graph-backing that #2712 deferred
(#883) and the “unbacked filename string” ADR &lt;code dir=&quot;auto&quot;&gt;2026-07-15-1&lt;/code&gt; flags at Claim 4.&lt;/p&gt;
&lt;h3 id=&quot;d4--recommended_role--recommended_model_tier-are-advisory-doctrine-offers-not-routing-truth&quot;&gt;D4 — &lt;code dir=&quot;auto&quot;&gt;recommended_role&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;recommended_model_tier&lt;/code&gt; are advisory doctrine &lt;strong&gt;offers&lt;/strong&gt;, not routing truth&lt;/h3&gt;
&lt;p&gt;Routing (which agent, which model) remains a charter/runtime decision (the “doctrine offers / charter activates
/ runtime consumes” spine). Doctrine offers a hint; the runtime/charter retains override authority. The design
MUST define the override seam and switch consumers to read the step grain — role/model is authored in ≥4 places
today (WP frontmatter, &lt;code dir=&quot;auto&quot;&gt;MissionStep.agent_profile&lt;/code&gt;, governance-profile, action-index); consolidation is a win
only if it becomes the single source, not a fifth parallel authority.&lt;/p&gt;
&lt;h3 id=&quot;d5--retain-the-governing-doctrine-scope-edges&quot;&gt;D5 — retain the governing-doctrine &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt; edges&lt;/h3&gt;
&lt;p&gt;The step schema keeps the &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt; (governing directives/tactics/procedures) edge population — the load-bearing
thing action nodes do today. It is not replaced by &lt;code dir=&quot;auto&quot;&gt;template&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;prompt&lt;/code&gt;.&lt;/p&gt;
&lt;h3 id=&quot;d6--mission_step_contract-is-a-steps-typed-io&quot;&gt;D6 — &lt;code dir=&quot;auto&quot;&gt;MISSION_STEP_CONTRACT&lt;/code&gt; is a step’s typed I/O&lt;/h3&gt;
&lt;p&gt;The declared-but-empty &lt;code dir=&quot;auto&quot;&gt;NodeKind.MISSION_STEP_CONTRACT&lt;/code&gt; becomes a step’s input/output signature, modelled as a
&lt;code dir=&quot;auto&quot;&gt;step → mission_step_contract&lt;/code&gt; edge (so contracts stay independently reusable/activatable).&lt;/p&gt;
&lt;h2 id=&quot;slices-prioritized-by-the-operator&quot;&gt;Slices (prioritized by the operator)&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Prio-0 (now):&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;S-A — Rename / vocabulary.&lt;/strong&gt; Narrow “template” to the artefact kind; establish “step” as the mission-type
structural term across docs, schema comments, and public surfaces. Low code, high clarity; unblocks the rest.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;S-B — Step authority / alignment.&lt;/strong&gt; Collapse the split-brain: make &lt;code dir=&quot;auto&quot;&gt;step.yaml&lt;/code&gt; (the &lt;code dir=&quot;auto&quot;&gt;MissionStep&lt;/code&gt; tree) the
&lt;strong&gt;authority&lt;/strong&gt;; derive &lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt; (and &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt;) as &lt;em&gt;projections&lt;/em&gt;; teach the DRG extractor to emit
from the step authority; switch consumers to the step grain (materialize WP role/model as caches, per D4).
Add &lt;code dir=&quot;auto&quot;&gt;recommended_model_tier&lt;/code&gt; (net-new; authored nowhere today) and the &lt;code dir=&quot;auto&quot;&gt;template&lt;/code&gt; reference to the step
schema.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;S-C — Fix missing spec/plan exemplars.&lt;/strong&gt; Author the content templates the three default types lack, and
graph-back them as &lt;code dir=&quot;auto&quot;&gt;step → template&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;instantiates&lt;/code&gt; edges (D3) — mint the &lt;code dir=&quot;auto&quot;&gt;spec&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;plan&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;task-prompt&lt;/code&gt;
scaffolds as &lt;code dir=&quot;auto&quot;&gt;template:&lt;/code&gt; nodes and teach the resolver to resolve-by-URN. This closes the #2689 uncreatable
regression &lt;strong&gt;and&lt;/strong&gt; #2712/#883. Each type’s templates hang on &lt;em&gt;its own&lt;/em&gt; step names (&lt;code dir=&quot;auto&quot;&gt;research&lt;/code&gt;:
&lt;code dir=&quot;auto&quot;&gt;scoping→methodology→…&lt;/code&gt;; &lt;code dir=&quot;auto&quot;&gt;documentation&lt;/code&gt;: &lt;code dir=&quot;auto&quot;&gt;discover→audit→…&lt;/code&gt;) — do not assume a &lt;code dir=&quot;auto&quot;&gt;specify&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;plan&lt;/code&gt; shape.
&lt;strong&gt;Re-scoped 2026-07-17 (see Amendment below):&lt;/strong&gt; S-C also &lt;strong&gt;retires the persisted &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; field&lt;/strong&gt;,
reading the step authority throughout (tidy-first, before authoring); &lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt; symmetry is split
to a distinct follow-up slice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Deferred (lower priority):&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;S-D — Substeps (recursion).&lt;/strong&gt; Add a &lt;code dir=&quot;auto&quot;&gt;DECOMPOSES_INTO&lt;/code&gt; relation + a new validator acyclicity DFS (the
validator guards only &lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;specializes_from&lt;/code&gt; today) + nested step URNs. Declare the stop-rule:
&lt;strong&gt;substeps = doctrine-authored/static; WPs = mission-emitted/dynamic&lt;/strong&gt; — never both for the same decomposition.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;S-E — Guards.&lt;/strong&gt; New &lt;code dir=&quot;auto&quot;&gt;NodeKind.GUARD&lt;/code&gt; + a &lt;code dir=&quot;auto&quot;&gt;GATES&lt;/code&gt; relation (net-new artefact modelling; guards are
engine-baked condition strings today). Separate epic slice.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;consequences&quot;&gt;Consequences&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Positive.&lt;/strong&gt; One discriminator (steps), one template vocabulary (the artefact kind), one step authority
(&lt;code dir=&quot;auto&quot;&gt;step.yaml&lt;/code&gt;). The &lt;code dir=&quot;auto&quot;&gt;mission_type → step → template&lt;/code&gt; chain becomes fully graph-backed and activation-gated,
extending #2712 and subsuming #2689’s &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; as a transitional read-projection (no rework of either).
The step model is ~70% shipped, so S-A/S-B/S-C are &lt;em&gt;promote + graph-back&lt;/em&gt;, not greenfield.
&lt;strong&gt;Superseded 2026-07-17 (see Amendment):&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; is not retained even as a persisted projection —
S-C retires the field and computes the projection at the consumption boundary from the step authority.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Costs / risks.&lt;/strong&gt; ~80 files / ~8 load-bearing seams. The two boundaries that must not be hand-waved into one
PR: the &lt;strong&gt;filesystem↔URN template duality&lt;/strong&gt; (S-C — a compatibility contract, resolve-by-name vs resolve-by-URN)
and the &lt;strong&gt;consumer read-source switch&lt;/strong&gt; for role/model (S-B — else a new parallel authority). Every new
NodeKind trips the golden-count + &lt;code dir=&quot;auto&quot;&gt;_ARCH_SHARD_N_FILES&lt;/code&gt; arch markers (mechanical). The #2660 meta-less /
fail-closed path must be preserved at every slice (do not reopen software-dev inference).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Explicitly NOT fixed by the schema alone.&lt;/strong&gt; The #2689 uncreatable regression is closed only by S-C
&lt;em&gt;authoring the missing content&lt;/em&gt;, not by the rename or authority collapse.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Deferred without debt.&lt;/strong&gt; Substeps and guards are net-new capabilities with their own DRG primitives; both
ride the existing activation filter when added. Nothing in S-A/S-B/S-C forecloses them.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;amendment--2026-07-17-s-c-re-scope-retire-template_set-as-a-persisted-field&quot;&gt;Amendment — 2026-07-17: S-C re-scope (retire &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; as a persisted field)&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Status:&lt;/strong&gt; Accepted (operator direction change; validated by a 3-lens research squad + a 3-lens adversarial
verification squad, all LAND). Mission: &lt;code dir=&quot;auto&quot;&gt;mission-step-creatability-01KXQA6R&lt;/code&gt; (#2724).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;What changed.&lt;/strong&gt; The original decision kept &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; as a &lt;strong&gt;transitional projected read-field&lt;/strong&gt; on
&lt;code dir=&quot;auto&quot;&gt;MissionType&lt;/code&gt; (“reuse the existing mechanism to limit blast radius”). On reflection this was judged a &lt;strong&gt;partial
cutover&lt;/strong&gt;: consumption still read a flattened dict instead of the step authority, it perpetuated a &lt;strong&gt;model
collision&lt;/strong&gt; with the unrelated scalar &lt;code dir=&quot;auto&quot;&gt;doctrine.template_set&lt;/code&gt; (separated only by the C-008 prose fence), and —
decisively — the still-authorable &lt;code dir=&quot;auto&quot;&gt;MissionType.template_set&lt;/code&gt; field plus the &lt;code dir=&quot;auto&quot;&gt;raw.get(&quot;template_set&quot;)&lt;/code&gt; fallback
(&lt;code dir=&quot;auto&quot;&gt;mission_type_repository.py:200-201&lt;/code&gt;) left a &lt;strong&gt;live split-brain vector&lt;/strong&gt;: a community/pack-authored &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt;
would be silently honored. All four concerns were confirmed in code.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;New direction.&lt;/strong&gt; S-C &lt;strong&gt;retires the persisted &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; field&lt;/strong&gt; and deletes the raw-YAML fallback; the
single behavioral consumer (&lt;code dir=&quot;auto&quot;&gt;resolve_configured_template&lt;/code&gt;) reads the lazy &lt;code dir=&quot;auto&quot;&gt;ResolvedMissionType.template_set&lt;/code&gt;
&lt;code dir=&quot;auto&quot;&gt;@cached_property&lt;/code&gt;, whose thunk now computes &lt;code dir=&quot;auto&quot;&gt;project_template_set(steps)&lt;/code&gt; from the step authority
(&lt;code dir=&quot;auto&quot;&gt;MissionStepRepository.resolve_all_for_mission_type(type, pack_context=None)&lt;/code&gt;, cached). Sequenced &lt;strong&gt;tidy-first&lt;/strong&gt;:
the cutover lands before the three types’ content is authored, so they are born on the clean surface and can never
carry the split-brain. A future pack authoring &lt;code dir=&quot;auto&quot;&gt;template_set:&lt;/code&gt; must &lt;strong&gt;fail loudly&lt;/strong&gt;, not be silently honored or
dropped.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Verified (adversarial squad).&lt;/strong&gt; Behavior-preserving — proven byte-for-byte at runtime (&lt;code dir=&quot;auto&quot;&gt;stored == project_template_set(steps)&lt;/code&gt; for all four types); no &lt;code dir=&quot;auto&quot;&gt;mission_types/*.yaml&lt;/code&gt; authors &lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt;, so the fallback
is already dead code. Blast radius ~4-5 code sites + ~6 field-pin test files; the resolver signature and the
#2689/#2660 fail-closed guards are untouched; the enduring &lt;code dir=&quot;auto&quot;&gt;TestSoftwareDevProjectionParity&lt;/code&gt; guard is retained
(only the &lt;code dir=&quot;auto&quot;&gt;MissionTypeRepository&lt;/code&gt;-injection half is disposable). The &lt;code dir=&quot;auto&quot;&gt;ResolvedMissionType.template_set&lt;/code&gt; property is
&lt;strong&gt;not&lt;/strong&gt; renamed (wide, all-green, no correctness benefit).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Split off.&lt;/strong&gt; The identical persisted-projection pattern for &lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;action_sequence&lt;/code&gt;&lt;/strong&gt; is &lt;strong&gt;not&lt;/strong&gt; folded in — it
crosses a categorically different gate (DRG shipped-graph + NFR-002 freshness), a hard non-empty
&lt;code dir=&quot;auto&quot;&gt;validate_action_sequence&lt;/code&gt; invariant, and the eager FSM hot-path — and is filed as a distinct follow-up slice
under #2721.&lt;/p&gt;
&lt;h2 id=&quot;references&quot;&gt;References&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;docs/plans/investigations/mission-type-step-model-unification.md&lt;/code&gt; (proposal + full grounding).&lt;/li&gt;
&lt;li&gt;ADR &lt;code dir=&quot;auto&quot;&gt;2026-07-15-1-doctrine-offers-charter-activates-runtime-consumes.md&lt;/code&gt; (Claim 4, S0/S4).&lt;/li&gt;
&lt;li&gt;Merged: #2712 (mission_type DRG node + mission_type→action edges, graph sharding), #2689 (&lt;code dir=&quot;auto&quot;&gt;template_set&lt;/code&gt; slot).&lt;/li&gt;
&lt;li&gt;Deferred edge class: #883 (mission_type→template). Epic: #2652.&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
