<?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/2026-07-21-1-in-tension-with-drg-edge/">
  <id>https://docs.spec-kitty.ai/adr/3.x/2026-07-21-1-in-tension-with-drg-edge/</id>
  <title>in_tension_with and reconciles_tension DRG edges (retiring opposed_by)</title>
  <updated>2026-07-21T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/3.x/2026-07-21-1-in-tension-with-drg-edge/"/>
  <summary>Retires `opposed_by` for symmetric, non-transitive `in_tension_with` plus `reconciles_tension` DRG relations, so tension is one traversable graph edge.</summary>
  <content type="html">&lt;span id=&quot;in_tension_with-and-reconciles_tension-drg-edges-retiring-opposed_by&quot;&gt;&lt;/span&gt;
&lt;p&gt;&lt;strong&gt;Filename:&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;2026-07-21-1-in-tension-with-drg-edge.md&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Status:&lt;/strong&gt; Superseded by &lt;a href=&quot;/adr/3.x/2026-07-26-3-impacts-edge-subsumes-in-tension-with/&quot;&gt;ADR 2026-07-26-3&lt;/a&gt; (2026-07-26)&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Do not implement from this ADR.&lt;/strong&gt; A signed &lt;code dir=&quot;auto&quot;&gt;impacts&lt;/code&gt; edge subsumes &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;:
&lt;code dir=&quot;auto&quot;&gt;impacts &amp;#x3C; 0&lt;/code&gt; &lt;strong&gt;is&lt;/strong&gt; a tension claim, so the boolean relation retires and a new
&lt;code dir=&quot;auto&quot;&gt;Relation.IMPACTS&lt;/code&gt; carries the value. &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; survives, re-pointed at
negative-&lt;code dir=&quot;auto&quot;&gt;impacts&lt;/code&gt; pairs. The symmetric single-edge storage convention described below
retires with the boolean; symmetric tensions become directed pairs. Kept for history.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;&lt;strong&gt;Date:&lt;/strong&gt; 2026-07-21&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Deciders:&lt;/strong&gt; Stijn Dejongh (operator), Architect Alphonso; input from Lynn Cole (#2537/#2538)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Technical Story:&lt;/strong&gt; &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2537&quot;&gt;#2537&lt;/a&gt; (supersedes the &lt;code dir=&quot;auto&quot;&gt;at_tension_with&lt;/code&gt; name proposed there), experiment &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2538&quot;&gt;#2538&lt;/a&gt;, epic &lt;a href=&quot;https://github.com/Priivacy-ai/spec-kitty/issues/2466&quot;&gt;#2466&lt;/a&gt;; related ADR &lt;a href=&quot;/adr/3.x/2026-07-18-1-charter-yaml-authoring-authority-and-extractor-retirement/&quot;&gt;2026-07-18-1&lt;/a&gt; (charter.yaml authoring authority + extractor retirement).&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id=&quot;context-and-problem-statement&quot;&gt;Context and Problem Statement&lt;/h2&gt;
&lt;p&gt;The Doctrine Reference Graph (DRG) models &lt;strong&gt;obligation&lt;/strong&gt; in abundance and &lt;strong&gt;tension&lt;/strong&gt; almost not at all. Measured on the current built-in layer (&lt;code dir=&quot;auto&quot;&gt;docs/plans/engineering-notes/doctrine-drg-missing-links-analysis.md&lt;/code&gt;): 255 &lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt; + 330 &lt;code dir=&quot;auto&quot;&gt;suggests&lt;/code&gt; edges, and &lt;strong&gt;zero&lt;/strong&gt; edges that say “these two rules compete.” When two &lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt;d directives land in the same activated context with no arbitration, the competition is invisible to the runtime and the agent silently picks which one wins — the “premature victory / quiet deferment” problem #2537 traces.&lt;/p&gt;
&lt;p&gt;The only machine-readable “these two are at odds” signal today is the &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; field, and it is broken three ways:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;It is mis-encoded in the graph.&lt;/strong&gt; The extractor (&lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/migration/extractor.py:373-391, 484-494&lt;/code&gt;) maps every &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; entry to &lt;code dir=&quot;auto&quot;&gt;Relation.REPLACES&lt;/code&gt;. &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; is a directional supersession relation; tension is neither directional nor supersession. The live result in &lt;code dir=&quot;auto&quot;&gt;src/doctrine/directive.graph.yaml&lt;/code&gt; is a nonsensical mutual cycle: &lt;code dir=&quot;auto&quot;&gt;DIRECTIVE_024 --replaces--&gt; DIRECTIVE_025&lt;/code&gt; &lt;strong&gt;and&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;DIRECTIVE_025 --replaces--&gt; DIRECTIVE_024&lt;/code&gt;. Two co-valid rules cannot each replace the other.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;It is consumed inconsistently.&lt;/strong&gt; The extractor emits &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; edges for directives and paradigms but &lt;strong&gt;not&lt;/strong&gt; for tactics — the tactic block only walks &lt;code dir=&quot;auto&quot;&gt;references&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;steps&lt;/code&gt;. So &lt;code dir=&quot;auto&quot;&gt;tactics/built-in/change-apply-smallest-viable-diff.tactic.yaml&lt;/code&gt;’s &lt;code dir=&quot;auto&quot;&gt;opposed_by: DIRECTIVE_025&lt;/code&gt; produces no edge at all. The field is dead for one of the three kinds that declare it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Nothing reads it for its stated purpose.&lt;/strong&gt; No activation-time or consistency-check path inspects &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; or any tension signal. &lt;code dir=&quot;auto&quot;&gt;charter pack consistency-check&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;src/charter/consistency_check.py&lt;/code&gt;) checks reference parity, kind violations, and duplicates — never tension. The lint &lt;code dir=&quot;auto&quot;&gt;ContradictionChecker&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/charter_runtime/lint/checks/contradiction.py&lt;/code&gt;) checks ADR-topic clashes and duplicate glossary senses — unrelated to &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;. The reconciliation prose that already exists inside &lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt;’s &lt;code dir=&quot;auto&quot;&gt;opposed_by.reason&lt;/code&gt; never reaches an agent as structure.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Separately, &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; is &lt;strong&gt;semantically overloaded&lt;/strong&gt;. Its six live usages split into two different concepts:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Genuine tension between two co-valid, activatable artefacts:&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;024 ↔ 025&lt;/code&gt; (locality vs boy-scout) and &lt;code dir=&quot;auto&quot;&gt;change-apply-smallest-viable-diff ↔ DIRECTIVE_025&lt;/code&gt;. Both sides are real doctrine, both stay valid, both can be simultaneously active.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Rejection of a named anti-pattern:&lt;/strong&gt; the three paradigm usages (&lt;code dir=&quot;auto&quot;&gt;domain-driven-design&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;brownfield-onboarding&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;c4-incremental-detail-modeling&lt;/code&gt;) point &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; at anti-patterns — &lt;code dir=&quot;auto&quot;&gt;anemic-domain-model&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;big-ball-of-mud&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;big-upfront-design&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;code-is-the-documentation&lt;/code&gt;, plus the tactics &lt;code dir=&quot;auto&quot;&gt;database-driven-design&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;single-diagram-architecture&lt;/code&gt;. The anti-pattern targets are &lt;strong&gt;not real doctrine artefacts&lt;/strong&gt; — they exist only as phantom nodes the extractor invents in &lt;code dir=&quot;auto&quot;&gt;paradigm.graph.yaml&lt;/code&gt; to hang the (mis-labelled) &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edge on. This is “this good approach supersedes that bad one,” which is closer to &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; than to tension.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Finally, the codebase is mid-migration &lt;strong&gt;away from field-authored relationships toward authored DRG edges&lt;/strong&gt;: WP06/WP07 already moved &lt;code dir=&quot;auto&quot;&gt;specializes-from&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;enhances&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;overrides&lt;/code&gt; from artifact fields to typed edges, guarded by &lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relationship_migration.py::test_built_in_relationships_authored_in_drg&lt;/code&gt;, which asserts &lt;strong&gt;no&lt;/strong&gt; field-authored relationships remain. ADR 2026-07-18-1 retired the extractor as authoring authority. &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; is the last field-form relationship holdout, and it points the wrong way against the established direction of travel.&lt;/p&gt;
&lt;h2 id=&quot;decision-drivers&quot;&gt;Decision Drivers&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Make doctrine competition &lt;strong&gt;queryable structure&lt;/strong&gt;, not prose buried in 18 artifact bodies.&lt;/li&gt;
&lt;li&gt;Encode tension with correct semantics (symmetric, non-supersession) instead of borrowing &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Give &lt;code dir=&quot;auto&quot;&gt;charter activate&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;charter pack consistency-check&lt;/code&gt; a way to &lt;strong&gt;warn&lt;/strong&gt; the operator when co-activated artefacts are in tension, with a concrete resolution.&lt;/li&gt;
&lt;li&gt;Stay on the field→edge migration path (WP06/WP07 precedent, extractor retirement) rather than entrenching a field form.&lt;/li&gt;
&lt;li&gt;Do not force the anti-pattern-rejection usages into a tension shape they do not fit.&lt;/li&gt;
&lt;li&gt;Warn, do not hard-block (Lynn’s explicit constraint in #2537).&lt;/li&gt;
&lt;li&gt;Every relation is self-describing, with the glossary holding the same definition (single canonical authority — no drift between code and docs).&lt;/li&gt;
&lt;li&gt;Cascade / multi-tier activation must reach every doctrine artefact kind, so tension between any co-activated kinds can actually surface.&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 — First-class symmetric &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; DRG edge + &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edge, remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;.&lt;/strong&gt; (Chosen.)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option B — Keep &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; as a field, but wire an activation-time reader for it.&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option C — Add the tension edge but keep &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; too (dual representation during a deprecation window).&lt;/strong&gt;&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 A”&lt;/strong&gt;, because it fixes the semantics, aligns with the field→edge migration already in flight, and gives the activation and consistency-check surfaces a single graph relation to traverse. The operator decisions below (Decisions 1–6, plus a related cross-cutting directive on cascade coverage) are the binding content of this ADR.&lt;/p&gt;
&lt;h3 id=&quot;decision-1--relation-name-is-in_tension_with&quot;&gt;Decision 1 — Relation name is &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;&lt;/h3&gt;
&lt;p&gt;Add &lt;code dir=&quot;auto&quot;&gt;IN_TENSION_WITH = &quot;in_tension_with&quot;&lt;/code&gt; to &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/models.py&lt;/code&gt;). This supersedes the &lt;code dir=&quot;auto&quot;&gt;at_tension_with&lt;/code&gt; name proposed in #2537 (grammatically correct: “A is in tension with B”). Feasible with no ripple: &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; is a plain &lt;code dir=&quot;auto&quot;&gt;StrEnum&lt;/code&gt; consumed by string comparison; there are &lt;strong&gt;no exhaustive &lt;code dir=&quot;auto&quot;&gt;match&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;case&lt;/code&gt; statements over &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt;&lt;/strong&gt; anywhere in &lt;code dir=&quot;auto&quot;&gt;src/&lt;/code&gt;, so a new member breaks no exhaustiveness contract. &lt;code dir=&quot;auto&quot;&gt;graph.yaml&lt;/code&gt; fragments validate against the enum via Pydantic, so the member must exist before any fragment authors the relation.&lt;/p&gt;
&lt;h3 id=&quot;decision-2--in_tension_with-is-symmetric-not-transitive&quot;&gt;Decision 2 — &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; is symmetric, NOT transitive&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Symmetric (bidirectional):&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;A in_tension_with B&lt;/code&gt; implies &lt;code dir=&quot;auto&quot;&gt;B in_tension_with A&lt;/code&gt;. (The operator’s note “A→B means A→C” is a typo for the symmetric closure “A→B ⇒ B→A”; the corrected semantics are stated here explicitly.)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Not transitive:&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;A⋈B&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;B⋈C&lt;/code&gt; do &lt;strong&gt;NOT&lt;/strong&gt; imply &lt;code dir=&quot;auto&quot;&gt;A⋈C&lt;/code&gt;. Transitivity is never computed; the checker inspects only declared pairs.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Storage recommendation — store one canonical edge, treat symmetrically at query time.&lt;/strong&gt; DRG edges are directed (&lt;code dir=&quot;auto&quot;&gt;source&lt;/code&gt;→&lt;code dir=&quot;auto&quot;&gt;target&lt;/code&gt;) and every existing consumer, including the cascade forward-closure, assumes direction. The DRG has no symmetric relation today. The two viable encodings are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;em&gt;(a) Store one edge, query both directions.&lt;/em&gt; Author a single &lt;code dir=&quot;auto&quot;&gt;A --in_tension_with--&gt; B&lt;/code&gt;. The tension checker builds the undirected view by unioning &lt;code dir=&quot;auto&quot;&gt;graph.edges_from(urn, IN_TENSION_WITH)&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;graph.edges_to(urn, IN_TENSION_WITH)&lt;/code&gt; (both helpers already exist on &lt;code dir=&quot;auto&quot;&gt;DRGGraph&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;&lt;em&gt;(b) Materialize both directions.&lt;/em&gt; Author &lt;code dir=&quot;auto&quot;&gt;A→B&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;B→A&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Recommend (a).&lt;/strong&gt; It keeps authoring DRY and — critically — removes a live data-integrity smell: today’s &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; pair carries &lt;strong&gt;two divergent &lt;code dir=&quot;auto&quot;&gt;reason&lt;/code&gt; texts&lt;/strong&gt; pointing opposite ways, which can (and did) drift. One canonical edge has one canonical reason. Materializing both directions doubles the authoring surface and invites exactly that drift, and it buys nothing because cascade must not follow this relation anyway (Decision 3). Canonical direction: author from the lexicographically-smaller URN to the larger, so the pair is deterministic and de-duplicable.&lt;/p&gt;
&lt;h3 id=&quot;decision-3--activation-and-consistency-check-flag-co-activated-tension-two-resolution-paths&quot;&gt;Decision 3 — Activation and consistency-check FLAG co-activated tension; two resolution paths&lt;/h3&gt;
&lt;p&gt;When two artefacts that are &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; each other are both active (or both about to be activated), the tooling raises an operator-facing finding. A tension pair &lt;code dir=&quot;auto&quot;&gt;(A, B)&lt;/code&gt; is treated as &lt;strong&gt;resolved&lt;/strong&gt; iff &lt;strong&gt;either&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;(a) Non-activation of one side&lt;/strong&gt; — A and B are not both in the active set; or&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;(b) Reconciliation artefact(s) bridge both sides&lt;/strong&gt; — each side has an &lt;strong&gt;active&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edge pointing at it. This is the implemented general rule: the checker tests &lt;strong&gt;per-side URN membership&lt;/strong&gt; in the set of reconciled targets (any active artefact with a &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edge into that URN counts), not “does a single artefact &lt;code dir=&quot;auto&quot;&gt;R&lt;/code&gt; carry edges to both.” The common case is one active artefact &lt;code dir=&quot;auto&quot;&gt;R&lt;/code&gt; with edges &lt;code dir=&quot;auto&quot;&gt;R --reconciles_tension--&gt; A&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;R --reconciles_tension--&gt; B&lt;/code&gt;; the rule is equally satisfied by two &lt;strong&gt;distinct&lt;/strong&gt; active artefacts &lt;code dir=&quot;auto&quot;&gt;R1 --reconciles_tension--&gt; A&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;R2 --reconciles_tension--&gt; B&lt;/code&gt;, each bridging one side. &lt;code dir=&quot;auto&quot;&gt;R&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;R1&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;R2&lt;/code&gt; are local (or built-in) doctrine artefacts whose bodies carry the reconciliation guidance; the &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edge(s) are how the graph records that the adjudication reaches this specific pair — a single reconciler spanning both sides is a special case of per-side bridging, not a separate, stricter requirement.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Add &lt;code dir=&quot;auto&quot;&gt;RECONCILES_TENSION = &quot;reconciles_tension&quot;&lt;/code&gt; to &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt;. A pair is resolved by (b) only when &lt;strong&gt;each&lt;/strong&gt; side has at least one active &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edge pointing at it — a pair with only one side bridged (a half-reconciled pair, whichever artefact(s) are involved) is still flagged as unreconciled.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Where it hooks in:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;charter pack consistency-check&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;run_consistency_check&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;src/charter/consistency_check.py&lt;/code&gt;) is the natural home. It already loads the activation-filtered DRG (&lt;code dir=&quot;auto&quot;&gt;filter_graph_by_activation&lt;/code&gt;) and holds the per-kind activation set. Add a new finding category (e.g. &lt;code dir=&quot;auto&quot;&gt;tension_unreconciled&lt;/code&gt;) computed from the activated graph’s &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; pairs minus those satisfied by (a) or (b). Keep it advisory (see consequences) rather than flipping &lt;code dir=&quot;auto&quot;&gt;coherent&lt;/code&gt; to hard-fail by default.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;charter activate&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/cli/commands/charter/activate.py&lt;/code&gt;) surfaces the same check as a &lt;code dir=&quot;auto&quot;&gt;[yellow]Warning[/yellow]&lt;/code&gt; for the artefact being activated, alongside the existing no-cascade warning path.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cascade integration (&lt;code dir=&quot;auto&quot;&gt;REFERENCE_RELATIONS&lt;/code&gt;).&lt;/strong&gt; Both new relations are deliberately &lt;strong&gt;excluded&lt;/strong&gt; from &lt;code dir=&quot;auto&quot;&gt;charter.cascade.REFERENCE_RELATIONS&lt;/code&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; must &lt;strong&gt;not&lt;/strong&gt; cascade — activating &lt;code dir=&quot;auto&quot;&gt;A&lt;/code&gt; must never auto-activate its opponent &lt;code dir=&quot;auto&quot;&gt;B&lt;/code&gt;; that would manufacture the exact conflict we are trying to make visible.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; must &lt;strong&gt;not&lt;/strong&gt; cascade — a reconciler (&lt;code dir=&quot;auto&quot;&gt;R&lt;/code&gt;, or &lt;code dir=&quot;auto&quot;&gt;R1&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;R2&lt;/code&gt; when the two sides are bridged separately) reconciles &lt;code dir=&quot;auto&quot;&gt;A&lt;/code&gt; and/or &lt;code dir=&quot;auto&quot;&gt;B&lt;/code&gt; but does not &lt;em&gt;require&lt;/em&gt; them; the reconciliation is only meaningful when &lt;code dir=&quot;auto&quot;&gt;A&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;B&lt;/code&gt; are already both active, so activating a reconciler must not drag them in.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;So the cascade reference set stays &lt;code dir=&quot;auto&quot;&gt;{REQUIRES, SUGGESTS, REFINES}&lt;/code&gt;. The tension checker reads these two relations directly off the graph; it does not ride the cascade closure. This keeps the cascade engine’s “pure reachability, no per-kind logic” contract intact.&lt;/p&gt;
&lt;h3 id=&quot;decision-4--remove-opposed_by-superseded-by-the-edge-representation&quot;&gt;Decision 4 — Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;, superseded by the edge representation&lt;/h3&gt;
&lt;p&gt;&lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; predates the DRG, is less maintainable, is mis-encoded (Decision context), and fights the field→edge migration. Remove it. Genuine tensions become &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; edges; the anti-pattern-rejection usages become &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges (Decision 5).&lt;/p&gt;
&lt;h3 id=&quot;decision-5--rejects-relation-for-anti-pattern-rejection-anti-patternsmell-artefacts-carry-a-marker&quot;&gt;Decision 5 — &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; relation for anti-pattern rejection; anti-pattern/smell artefacts carry a marker&lt;/h3&gt;
&lt;p&gt;The paradigm &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; usages that point at anti-patterns (&lt;code dir=&quot;auto&quot;&gt;anemic-domain-model&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;big-ball-of-mud&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;big-upfront-design&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;code-is-the-documentation&lt;/code&gt;, plus tactic targets &lt;code dir=&quot;auto&quot;&gt;database-driven-design&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;single-diagram-architecture&lt;/code&gt;) are &lt;strong&gt;rejections of a bad approach&lt;/strong&gt;, not tension between co-valid peers. Add &lt;code dir=&quot;auto&quot;&gt;REJECTS = &quot;rejects&quot;&lt;/code&gt; to &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt;: a directional edge &lt;code dir=&quot;auto&quot;&gt;good-artefact --rejects--&gt; anti-pattern&lt;/code&gt;. This supersedes the earlier “keep them as &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt;” fallback — &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; implies the target was once a valid predecessor, which an anti-pattern never was. &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; is directional (not symmetric) and does &lt;strong&gt;not&lt;/strong&gt; cascade (excluded from &lt;code dir=&quot;auto&quot;&gt;REFERENCE_RELATIONS&lt;/code&gt; alongside the tension relations).&lt;/p&gt;
&lt;p&gt;The anti-pattern targets must be &lt;strong&gt;first-class, marked nodes&lt;/strong&gt;, not phantom nodes the extractor invents. Anti-pattern / smell artefacts carry an explicit marker — a &lt;code dir=&quot;auto&quot;&gt;tags&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;labels&lt;/code&gt; attribute (e.g. &lt;code dir=&quot;auto&quot;&gt;tags: [anti-pattern]&lt;/code&gt; or &lt;code dir=&quot;auto&quot;&gt;[smell]&lt;/code&gt;) — so consumers, the rejection/tension checkers, and rendered agent context can distinguish “avoid this” targets from ordinary doctrine. This makes &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; validatable (a &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edge must terminate at a marked anti-pattern/smell node) and keeps the graph honest about which nodes represent things to avoid rather than things to do.&lt;/p&gt;
&lt;h3 id=&quot;decision-6--every-relation-member-is-self-describing-with-glossary-parity&quot;&gt;Decision 6 — Every &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; member is self-describing, with glossary parity&lt;/h3&gt;
&lt;p&gt;Each &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; enum member MUST carry a human-readable description of its meaning at the point of definition (&lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/models.py&lt;/code&gt;), and the project glossary MUST hold the same definition. The three members introduced by this ADR — &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; — each get a description in the enum and a matching glossary entry; this is also a standing rule applied to existing relations going forward. Single canonical authority (charter principle): the enum description and the glossary definition are the same text (or, for the human-facing &lt;code dir=&quot;auto&quot;&gt;docs/context&lt;/code&gt; glossary, a definition drawn from it), so relation semantics cannot drift between code and docs.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actual enforcement surfaces (as implemented):&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;RELATION_DESCRIPTIONS&lt;/code&gt; in &lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/models.py&lt;/code&gt; is the single canonical authority for the description text. &lt;code dir=&quot;auto&quot;&gt;docs/architecture/doctrine-relationships.md&lt;/code&gt; (“Tension vocabulary” section) mirrors that text verbatim for human readers, and &lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relation_doc_parity.py&lt;/code&gt; is a content-equality test that fails red — naming the diverging relation — the moment that mirror drifts from the registry. The human-facing project glossary (&lt;code dir=&quot;auto&quot;&gt;docs/context/charter.md&lt;/code&gt;) additionally carries &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, and &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; as glossary entries, with definitions drawn from the registry text; the other nine pre-existing &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; members (&lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;suggests&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;refines&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;specializes_from&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;enhances&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;overrides&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;applies&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;vocabulary&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;instantiates&lt;/code&gt;) do not yet have individual &lt;code dir=&quot;auto&quot;&gt;docs/context&lt;/code&gt; glossary entries — backfilling those is out of scope here (spec.md Assumption A2) and is tracked as follow-up. There is no separate “terminology guard / glossary integrity pipeline” test wired to this parity obligation today; &lt;code dir=&quot;auto&quot;&gt;test_relation_doc_parity.py&lt;/code&gt; (scoped to &lt;code dir=&quot;auto&quot;&gt;doctrine-relationships.md&lt;/code&gt;) is the concrete, executable guard, and the &lt;code dir=&quot;auto&quot;&gt;docs/context/charter.md&lt;/code&gt; entries are presently maintained by convention rather than an automated content-equality check.&lt;/p&gt;
&lt;h3 id=&quot;related-operator-directive-broader-than-this-adr--cascade-applies-to-all-doctrine-artefact-kinds&quot;&gt;Related operator directive (broader than this ADR) — cascade applies to ALL doctrine artefact kinds&lt;/h3&gt;
&lt;p&gt;The operator directs that the cascade / multi-tier pack + charter-activation system apply to &lt;strong&gt;every&lt;/strong&gt; doctrine &lt;code dir=&quot;auto&quot;&gt;artefactKind&lt;/code&gt;, not just the subset reachable today. This is the substrate the tension check (Decision 3) depends on: if activation cannot co-activate scoped directives/tactics/templates across kinds, tension between them can never surface. The missing-links analysis on this branch shows cascade currently dead-ends — &lt;code dir=&quot;auto&quot;&gt;REFERENCE_RELATIONS&lt;/code&gt; reaches actions from a mission-type but not the directives/tactics/templates those actions &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;instantiate&lt;/code&gt;. Making cascade kind-complete is broader than tension and is recorded here as the governing directive; the concrete relation-coverage work is its own design/mission scope (candidate for a dedicated ADR under epic #2466). Interaction note: &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; remain &lt;strong&gt;excluded&lt;/strong&gt; from the cascade reference set even under kind-complete cascade — Decision 3’s rationale is unchanged.&lt;/p&gt;
&lt;h3 id=&quot;note--delegates_to-is-a-topological-relation-with-littleno-runtime-impact-today&quot;&gt;Note — &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; is a topological relation with little/no runtime impact today&lt;/h3&gt;
&lt;p&gt;&lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; is a doctrine relation (an authored edge between agent profiles) that currently carries essentially &lt;strong&gt;no runtime behaviour&lt;/strong&gt; — the missing-links analysis on this branch measured &lt;strong&gt;zero&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; edges in the built-in layer. It is recorded here so future readers do not mistake it for dead vocabulary. Its intended purpose is to enable &lt;strong&gt;loose-governance, self-organizing cooperation between agents&lt;/strong&gt;, as an alternative to running a full scripted mission. The model: the operator opens a limited-scope op and simply says “get it done”; the agents then &lt;strong&gt;swarm and self-steer along the &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; edges&lt;/strong&gt; rather than following a mission pipeline. Illustrative chain — Alphonso writes an ADR → a reviewer/writer agent reviews it → an implementation agent automates the change → the writer agent updates the docsite, each handoff following a &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; edge. Making &lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt; runtime-effective (a self-steering delegation loop) is future scope, independent of the tension edges decided here; the relation is documented so its intent survives until that work is scheduled.&lt;/p&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;Tension becomes queryable: one relation the activation and consistency surfaces can traverse, and one relation an agent’s composed context can render (“these two compete; here is how they reconcile”).&lt;/li&gt;
&lt;li&gt;The nonsensical mutual &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; cycle on &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; disappears; supersession semantics stop being borrowed for tension.&lt;/li&gt;
&lt;li&gt;The tactic tension (&lt;code dir=&quot;auto&quot;&gt;change-apply-smallest-viable-diff ↔ 025&lt;/code&gt;) that the extractor silently dropped is now representable — a previously-lost signal is recovered.&lt;/li&gt;
&lt;li&gt;Removing the last field-form relationship completes the WP06/WP07 migration direction and keeps &lt;code dir=&quot;auto&quot;&gt;test_relationship_migration.py&lt;/code&gt;’s “no field-authored relationships” invariant honest.&lt;/li&gt;
&lt;li&gt;One canonical edge = one canonical reason; the divergent-reason drift on &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; is eliminated.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;The built-in default pack ships &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; &lt;strong&gt;both active&lt;/strong&gt; and &lt;strong&gt;in tension&lt;/strong&gt;, so a naive hard-block would fail the default pack out of the box. Mitigation: ship a &lt;strong&gt;built-in reconciliation artefact&lt;/strong&gt; for the &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; (and &lt;code dir=&quot;auto&quot;&gt;smallest-viable-diff&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt;) pair — the reconciliation text already exists verbatim in &lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt;’s &lt;code dir=&quot;auto&quot;&gt;opposed_by.reason&lt;/code&gt; — and wire it with &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edges so the default pack is self-consistent. Until that artefact exists, the tension check must be advisory, not coherence-failing. This matches Lynn’s “warn, do not hard-block.”&lt;/li&gt;
&lt;li&gt;Removal touches schemas, models, the shared &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; model, the extractor, six built-in YAMLs, generated graph fragments, docs, and tests in one migration (blast radius below). The dead-symbol gate requires &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; and the &lt;code dir=&quot;auto&quot;&gt;contradiction&lt;/code&gt; schema definition to be removed once the last field usage is gone.&lt;/li&gt;
&lt;li&gt;Introduces the DRG’s first symmetric relation; every future tension consumer must remember to query both directions (documented in Decision 2).&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;The anti-pattern-rejection paradigm usages are re-classified to &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges, not deleted (Decision 5 / Migration). The phantom anti-pattern nodes are promoted to first-class marked nodes (&lt;code dir=&quot;auto&quot;&gt;tags: [anti-pattern]&lt;/code&gt;) rather than extractor-invented.&lt;/li&gt;
&lt;li&gt;No change to the cascade reference set — the three new relations stay excluded even under the kind-complete cascade directive.&lt;/li&gt;
&lt;li&gt;Cascade kind-completeness (related directive) is broader than this ADR and lands as separate scope; this ADR only records the directive and its dependency relationship to the tension check.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;confirmation&quot;&gt;Confirmation&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;Relation.IN_TENSION_WITH&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;Relation.RECONCILES_TENSION&lt;/code&gt;, and &lt;code dir=&quot;auto&quot;&gt;Relation.REJECTS&lt;/code&gt; exist, each with a description; &lt;code dir=&quot;auto&quot;&gt;graph.yaml&lt;/code&gt; fragments validate.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;docs/architecture/doctrine-relationships.md&lt;/code&gt; defines &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; with text matching the enum descriptions verbatim (Decision 6); &lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relation_doc_parity.py&lt;/code&gt; passes. &lt;code dir=&quot;auto&quot;&gt;docs/context/charter.md&lt;/code&gt; additionally carries a human-glossary entry for each of the three relations, consistent with the enum descriptions.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;grep -rn &quot;opposed_by&quot; src/&lt;/code&gt; returns no hits across the doctrine authoring surfaces after migration (field, schema, model, extractor, built-in YAML); the only remaining &lt;code dir=&quot;auto&quot;&gt;src/&lt;/code&gt; hits are the &lt;code dir=&quot;auto&quot;&gt;spec-kitty migrate rewrite-opposed-by&lt;/code&gt; subcommand (&lt;code dir=&quot;auto&quot;&gt;migration/rewrite_opposed_by.py&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;cli/commands/migrate_cmd.py&lt;/code&gt;), which names the legacy field it rewrites for downstream/org packs.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relationship_migration.py::test_built_in_relationships_authored_in_drg&lt;/code&gt; still passes (no field-authored relationships).&lt;/li&gt;
&lt;li&gt;Every &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edge terminates at a node marked &lt;code dir=&quot;auto&quot;&gt;tags: [anti-pattern]&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;[smell]&lt;/code&gt; (Decision 5) — a test asserts no &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edge points at an unmarked node.&lt;/li&gt;
&lt;li&gt;A focused test asserts: (i) &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; resolves to a single &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; edge, not a &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; cycle; (ii) an unreconciled active tension pair produces the new consistency finding; (iii) adding an active &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; artefact clears it; (iv) activating one side of a tension does &lt;strong&gt;not&lt;/strong&gt; cascade-activate the other; (v) the paradigm anti-pattern usages resolve to &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges, not &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Confidence: high on Decisions 1, 2, 4, 5, 6 (mechanical, precedented); medium on Decision 3’s default-pack self-consistency (depends on authoring the built-in reconciliation artefact before any hard-block is enabled). The cascade-all-kinds directive is out of scope for confirmation here.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;migration-plan-and-blast-radius&quot;&gt;Migration plan and blast radius&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Concept split governs the migration.&lt;/strong&gt; Only genuine two-active-artefact tensions become &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;. The paradigm anti-pattern rejections are a different concept and MUST NOT be forced into symmetric tension (doing so would falsely claim “big-ball-of-mud is in tension with DDD” and demand a reconciliation artefact for an anti-pattern). Their disposition (Decision 5): author them as directional &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges, and mark each anti-pattern/smell target node with a &lt;code dir=&quot;auto&quot;&gt;tags&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;labels: [anti-pattern]&lt;/code&gt; (or &lt;code dir=&quot;auto&quot;&gt;[smell]&lt;/code&gt;) attribute so it is a first-class node rather than a phantom the extractor invents.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; stays canonical — do not remove it.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; (and the pack-augmentation family &lt;code dir=&quot;auto&quot;&gt;enhances&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;overrides&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;specializes_from&lt;/code&gt;) are canonical relations whose primary role is &lt;strong&gt;multi-tier pack composition&lt;/strong&gt; (an org/project pack superseding or augmenting built-in doctrine). Spec Kitty does not currently set up its own multi-tier pack layering, so these relations are sparse-to-absent in the built-in layer &lt;strong&gt;by design&lt;/strong&gt; — that is expected, not a gap. This migration removes &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edges &lt;strong&gt;only&lt;/strong&gt; where they were mis-minted from &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; (the &lt;code dir=&quot;auto&quot;&gt;024↔025&lt;/code&gt; cycle and the paradigm anti-pattern usages); every other &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edge is left intact. The relations this ADR removes from the codebase entirely are the &lt;strong&gt;phantom lint-only relations &lt;code dir=&quot;auto&quot;&gt;governs&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;supersedes&lt;/code&gt;&lt;/strong&gt; (never enum members, never authorable — see the orphan-rule cleanup), not &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; edges to author (genuine tensions):&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;directive:DIRECTIVE_024 ↔ directive:DIRECTIVE_025&lt;/code&gt; — one canonical edge in &lt;code dir=&quot;auto&quot;&gt;src/doctrine/directive.graph.yaml&lt;/code&gt;, replacing the two mutual &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edges.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;tactic:change-apply-smallest-viable-diff ↔ directive:DIRECTIVE_025&lt;/code&gt; — one edge (previously never emitted).&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Blast radius (concrete file/impact list):&lt;/strong&gt;&lt;/p&gt;
































































































































































&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Area&lt;/th&gt;&lt;th&gt;File&lt;/th&gt;&lt;th&gt;Impact&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;Enum&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/models.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Add &lt;code dir=&quot;auto&quot;&gt;IN_TENSION_WITH&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;RECONCILES_TENSION&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;REJECTS&lt;/code&gt; to &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt;; each member carries a description (Decision 6); document symmetric-not-transitive for &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Doc parity&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;docs/architecture/doctrine-relationships.md&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relation_doc_parity.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Mirror the enum descriptions verbatim for &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt;; enforce content-equality parity against &lt;code dir=&quot;auto&quot;&gt;RELATION_DESCRIPTIONS&lt;/code&gt; with a dedicated test (Decision 6).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Glossary&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;docs/context/charter.md&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Add human-glossary entries for &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt;, consistent with the enum descriptions (Decision 6).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Node marker&lt;/td&gt;&lt;td&gt;anti-pattern/smell target nodes&lt;/td&gt;&lt;td&gt;Add &lt;code dir=&quot;auto&quot;&gt;tags&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;labels: [anti-pattern]&lt;/code&gt; (or &lt;code dir=&quot;auto&quot;&gt;[smell]&lt;/code&gt;) marker to anti-pattern nodes so &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; targets are first-class and validatable (Decision 5).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Schema&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/schemas/directive.schema.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; property + the &lt;code dir=&quot;auto&quot;&gt;contradiction&lt;/code&gt; definition.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Schema&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/schemas/tactic.schema.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; property + &lt;code dir=&quot;auto&quot;&gt;contradiction&lt;/code&gt; definition.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Schema&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/schemas/paradigm.schema.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; property + &lt;code dir=&quot;auto&quot;&gt;contradiction&lt;/code&gt; definition.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Model&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/directives/models.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; field + &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; import.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Model&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/tactics/models.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; field + &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; import.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Model&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/paradigms/models.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; field + &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; import.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Model&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/shared/models.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove now-dead &lt;code dir=&quot;auto&quot;&gt;Contradiction&lt;/code&gt; model + &lt;code dir=&quot;auto&quot;&gt;__all__&lt;/code&gt; entry (dead-symbol gate).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Extractor&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/drg/migration/extractor.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove both &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; blocks (directive ~373-391, paradigm ~484-494) that emit &lt;code dir=&quot;auto&quot;&gt;Relation.REPLACES&lt;/code&gt;. (No tactic block exists — the tactic field was never extracted.) Extractor is retired as authoring authority (ADR 2026-07-18-1) but the code path still runs; leaving the blocks would keep minting &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; from a removed field.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/directives/built-in/024-locality-of-change.directive.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/directives/built-in/025-boy-scout-rule.directive.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;; preserve its reconciliation text (source for the built-in reconciliation artefact).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/tactics/built-in/change-apply-smallest-viable-diff.tactic.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/paradigms/built-in/domain-driven-design.paradigm.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt;; its rejections re-express as authored &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/paradigms/built-in/brownfield-onboarding.paradigm.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Same.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Built-in YAML&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/paradigms/built-in/c4-incremental-detail-modeling.paradigm.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Same.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Generated graph&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/directive.graph.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Replace the two &lt;code dir=&quot;auto&quot;&gt;024↔025&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edges with one &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; edge.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Generated graph&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/paradigm.graph.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Author the paradigm rejection &lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt; edges; promote phantom anti-pattern targets to marked nodes (&lt;code dir=&quot;auto&quot;&gt;tags: [anti-pattern]&lt;/code&gt;).&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Generated graph&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/doctrine/tactic.graph.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Author the &lt;code dir=&quot;auto&quot;&gt;smallest-viable-diff → 025&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; edge.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;New built-in&lt;/td&gt;&lt;td&gt;reconciliation directive/tactic + &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt; edges&lt;/td&gt;&lt;td&gt;Ship a reconciliation artefact for the &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; (and smallest-diff/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt;) tension so the default pack is self-consistent under the check.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Orphan lint (#2737)&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/charter_runtime/lint/checks/orphan.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Remove the phantom &lt;code dir=&quot;auto&quot;&gt;governs&lt;/code&gt; (directive) and &lt;code dir=&quot;auto&quot;&gt;supersedes&lt;/code&gt; (adr) branches from &lt;code dir=&quot;auto&quot;&gt;_ORPHAN_RULES&lt;/code&gt; — neither is a &lt;code dir=&quot;auto&quot;&gt;Relation&lt;/code&gt; enum member, so neither is ever authorable; they misfire on 100% of directives. Retire those branches (or, for directives, re-point at the relations that actually reach them: &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;requires&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;suggests&lt;/code&gt;) so only the 2 genuinely-disconnected directives surface. Closes #2737.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Consistency check&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/charter/consistency_check.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Add the &lt;code dir=&quot;auto&quot;&gt;tension_unreconciled&lt;/code&gt; finding computed over the activation-filtered graph.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;CLI&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/cli/commands/charter/activate.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Surface the tension warning on activate.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Docs&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;docs/architecture/04_implementation_mapping/README.md&lt;/code&gt; (lines ~250, 260-263, 310-312, 335)&lt;/td&gt;&lt;td&gt;Replace &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; reference-direction + “Contradiction semantics” + schema-table + status-table entries with the edge model.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Docs&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;docs/guides/synthesize-doctrine.md&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Update the contradictions mention.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Tests&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_directive_consistency.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Drop/replace &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; assertions.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Tests&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;tests/doctrine/drg/migration/test_extractor.py&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Drop the &lt;code dir=&quot;auto&quot;&gt;opposed_by → replaces&lt;/code&gt; mapping test; add the new-edge expectations.&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Fixtures&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;tests/doctrine/fixtures/paradigm/valid/with-tactic-refs.yaml&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Drop &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; (schema &lt;code dir=&quot;auto&quot;&gt;additionalProperties: false&lt;/code&gt; will reject it).&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;&lt;strong&gt;Runtime-read safety confirmed.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;grep -rn &quot;opposed_by&quot; src/&lt;/code&gt; shows the only runtime &lt;em&gt;readers&lt;/em&gt; are the extractor blocks (→ &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt;). Nothing else reads the field: not the models beyond declaration, not &lt;code dir=&quot;auto&quot;&gt;consistency_check.py&lt;/code&gt;, not the lint &lt;code dir=&quot;auto&quot;&gt;ContradictionChecker&lt;/code&gt; (which is about ADR topics/glossary), not &lt;code dir=&quot;auto&quot;&gt;charter activate&lt;/code&gt;. Removing the field therefore silently breaks nothing at runtime — the only behavioral change is that the mis-encoded &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edges stop being minted.&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--symmetric-in_tension_with--reconciles_tension-remove-opposed_by-chosen&quot;&gt;Option A — symmetric &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt; + &lt;code dir=&quot;auto&quot;&gt;reconciles_tension&lt;/code&gt;, remove &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; (chosen)&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; correct semantics; queryable at activation/consistency-check; completes the field→edge migration; recovers the dropped tactic tension; one canonical reason per pair.
&lt;strong&gt;Cons:&lt;/strong&gt; first symmetric relation (consumers must query both directions); default pack needs a reconciliation artefact before any hard-block; multi-file migration touching the dead-symbol gate.&lt;/p&gt;
&lt;h3 id=&quot;option-b--keep-opposed_by-as-a-field-add-a-reader&quot;&gt;Option B — keep &lt;code dir=&quot;auto&quot;&gt;opposed_by&lt;/code&gt; as a field, add a reader&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; smaller diff; no enum change.
&lt;strong&gt;Cons:&lt;/strong&gt; entrenches the field form against the WP06/WP07 direction and ADR 2026-07-18-1; leaves the field consumed inconsistently across kinds (dead for tactics); the graph still lacks a tension relation, so cascade/lint/consistency all keep improvising; the mis-encoded &lt;code dir=&quot;auto&quot;&gt;replaces&lt;/code&gt; edges persist.&lt;/p&gt;
&lt;h3 id=&quot;option-c--dual-representation-edge--field-during-a-deprecation-window&quot;&gt;Option C — dual representation (edge + field) during a deprecation window&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; gradual migration.
&lt;strong&gt;Cons:&lt;/strong&gt; two sources of truth for the same fact invite drift (already observed in the &lt;code dir=&quot;auto&quot;&gt;024&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;025&lt;/code&gt; divergent reasons); the “no field-authored relationships” invariant cannot hold during the window; more surface to test and reconcile for no lasting benefit given the small, enumerable usage set (six files).&lt;/p&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Analysis: &lt;code dir=&quot;auto&quot;&gt;docs/plans/engineering-notes/doctrine-drg-missing-links-analysis.md&lt;/code&gt; (cascade dead-ends, relation-usage census).&lt;/li&gt;
&lt;li&gt;Migration precedent: &lt;code dir=&quot;auto&quot;&gt;tests/doctrine/test_relationship_migration.py&lt;/code&gt; (field→edge, zero-loss, no-field invariant).&lt;/li&gt;
&lt;li&gt;Cascade engine + reference set: &lt;code dir=&quot;auto&quot;&gt;src/charter/cascade.py&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;REFERENCE_RELATIONS&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;Consistency check surface: &lt;code dir=&quot;auto&quot;&gt;src/charter/consistency_check.py&lt;/code&gt;; CLI &lt;code dir=&quot;auto&quot;&gt;spec-kitty charter pack consistency-check&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Related: ADR 2026-07-18-1 (charter.yaml authoring authority, extractor retirement).&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
