<?xml version="1.0" encoding="utf-8"?>
<entry xmlns="http://www.w3.org/2005/Atom" xml:base="https://docs.spec-kitty.ai/plans/charter-resolution/program-brief/">
  <id>https://docs.spec-kitty.ai/plans/charter-resolution/program-brief/</id>
  <title>Charter-Resolution Program Brief</title>
  <updated>2026-09-30T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/plans/charter-resolution/program-brief/"/>
  <summary>Investigation-grounded brief for the charter-resolution program: root causes, mission decomposition, sequencing, risks and spec-time operator decisions.</summary>
  <content type="html">&lt;h2 id=&quot;mission-brief--doctrine-reach--resolution-honesty&quot;&gt;Mission Brief — Doctrine Reach &amp;#x26; Resolution Honesty&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Prepared from:&lt;/strong&gt; a 4-lens design investigation (load-path, cascade/edges, reach/delivery, validation-honesty) traced against current &lt;code dir=&quot;auto&quot;&gt;main&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;6ec05086f&lt;/code&gt;, post-#3520/#3534).
&lt;strong&gt;Rolls up to:&lt;/strong&gt; umbrella tracker &lt;strong&gt;#3530&lt;/strong&gt; (org-tier doctrine loads healthy but reaches no consumer) and fail-loud epic &lt;strong&gt;#3410&lt;/strong&gt;; reach sub-epic &lt;strong&gt;#3526&lt;/strong&gt;.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id=&quot;1-problem-statement-bluf&quot;&gt;1. Problem statement (BLUF)&lt;/h2&gt;
&lt;p&gt;Authored governance — org-pack and project-tier doctrine — &lt;strong&gt;silently fails to reach the dispatched agent.&lt;/strong&gt; Content loads, validates, reports healthy, then drops somewhere between authoring and consumption. The operator sees green checks and empty results. Concretely, on &lt;code dir=&quot;auto&quot;&gt;main&lt;/code&gt; as of 2026-08-19: activating any built-in mission type cascades to &lt;strong&gt;0&lt;/strong&gt; kinds; org packs authored in the canonical &lt;code dir=&quot;auto&quot;&gt;drg/fragment.yaml&lt;/code&gt; shape contribute &lt;strong&gt;nothing&lt;/strong&gt; to cascade; org/project doctrine of a given kind is under-loaded relative to built-in (measured &lt;strong&gt;71% tactic undercount&lt;/strong&gt;); glossary packs and 3 of 5 profile reference channels reach no agent; and several canonical facts are hand-restated in 5 places, two already drifted and failing open.&lt;/p&gt;
&lt;h2 id=&quot;2-root-causes-one-disease-six-faces&quot;&gt;2. Root causes (one disease, six faces)&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Org/project tiers are second-class vs built-in.&lt;/strong&gt; Built-in discovery &lt;code dir=&quot;auto&quot;&gt;rglob&lt;/code&gt;s unconditionally; org/project use non-recursive &lt;code dir=&quot;auto&quot;&gt;glob&lt;/code&gt;; and the &lt;em&gt;loader&lt;/em&gt; (&lt;code dir=&quot;auto&quot;&gt;doctrine/base.py&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;agent_profiles/repository.py&lt;/code&gt;) and the &lt;em&gt;charter-activation resolver&lt;/em&gt; (&lt;code dir=&quot;auto&quot;&gt;charter/kind_vocabulary.py&lt;/code&gt;) each decide recursion independently, so they disagree per kind. → &lt;strong&gt;#3490, #3426&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Two DRG read-paths never meet.&lt;/strong&gt; Cascade reads root-level &lt;code dir=&quot;auto&quot;&gt;*.graph.yaml&lt;/code&gt; via &lt;code dir=&quot;auto&quot;&gt;_drg_helpers.load_validated_graph&lt;/code&gt;; org &lt;code dir=&quot;auto&quot;&gt;drg/fragment.yaml&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;OrgDRGFragment&lt;/code&gt;) is read by a &lt;em&gt;separate&lt;/em&gt; path (&lt;code dir=&quot;auto&quot;&gt;org_pack_loader.load_org_pack → merge.merge_three_layers&lt;/code&gt;) consumed only by diagnostics. Nothing bridges fragment edges into cascade. → &lt;strong&gt;#3572&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Cascade traversal dead-ends.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;REFERENCE_RELATIONS = {requires, suggests, refines}&lt;/code&gt;; &lt;code dir=&quot;auto&quot;&gt;mission_type&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;action&lt;/code&gt; nodes carry &lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;instantiates&lt;/code&gt; edges that cascade never follows, so the forward closure reaches the action node and stops. → &lt;strong&gt;#2829&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Edges absent or unvalidated.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;collaboration.operating-procedures&lt;/code&gt; is a validated &lt;code dir=&quot;auto&quot;&gt;list[str]&lt;/code&gt; whose &lt;em&gt;values&lt;/em&gt; are never checked against nodes — measured &lt;strong&gt;36 of 50&lt;/strong&gt; name no node; real edges are hand-pinned per-profile; a handful of activated artefacts have zero inbound edges. → &lt;strong&gt;#2994, #3352, #3009&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Delivery tables/renderers have silent no-ops.&lt;/strong&gt; &lt;code dir=&quot;auto&quot;&gt;slot=None&lt;/code&gt; with no stated reason (glossary), &lt;code dir=&quot;auto&quot;&gt;body_fn=None&lt;/code&gt; pointer-only (styleguide/toolguide), a schema field no renderer reads (operating-procedures), a struct field permanently shadowed by a required sibling (procedure step &lt;code dir=&quot;auto&quot;&gt;description&lt;/code&gt;), and a builder that can’t resolve &lt;code dir=&quot;auto&quot;&gt;.kittify/agent_profiles&lt;/code&gt;. → &lt;strong&gt;#3488, #3489, #3176&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Canonical facts restated by hand / read through glob/name proxies, failing open.&lt;/strong&gt; Plural↔singular kind map in 5 sites (2 drifted); validator globs &lt;code dir=&quot;auto&quot;&gt;*.graph.yaml&lt;/code&gt; and misses &lt;code dir=&quot;auto&quot;&gt;fragment.yaml&lt;/code&gt;; &lt;code dir=&quot;auto&quot;&gt;context --json&lt;/code&gt; omits &lt;code dir=&quot;auto&quot;&gt;procedures[]&lt;/code&gt; that the text render ships; pack resolution is name-based instead of &lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt;-keyed. → &lt;strong&gt;#2981, #3573, #3389, #3574&lt;/strong&gt;&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;strong&gt;Meta-cause:&lt;/strong&gt; &lt;em&gt;the resolution surface restates canonical facts instead of deriving them from a single total authority, and every divergence degrades silently (fake-green).&lt;/em&gt; This is the #1868 “canonical seams in name only” thesis and the #3410 fail-loud thesis, expressed across charter resolution.&lt;/p&gt;
&lt;h2 id=&quot;3-proposed-approach-four-principles&quot;&gt;3. Proposed approach (four principles)&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Derive, don’t restate.&lt;/strong&gt; Single derived authorities (recursion policy; plural↔singular kind vocab) built from &lt;code dir=&quot;auto&quot;&gt;ArtifactKind&lt;/code&gt;; collapse the hand-copies. House precedents already on &lt;code dir=&quot;auto&quot;&gt;main&lt;/code&gt;: &lt;code dir=&quot;auto&quot;&gt;PROJECT_KIND_DIRS&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;ORG_PLURAL_TO_SINGULAR_KIND&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;_derive_plural_to_singular&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Bridge the read-paths.&lt;/strong&gt; Route org &lt;code dir=&quot;auto&quot;&gt;drg/fragment.yaml&lt;/code&gt; edges into the cascade graph via the &lt;em&gt;existing&lt;/em&gt; &lt;code dir=&quot;auto&quot;&gt;merge_three_layers&lt;/code&gt; machinery (it already owns endpoint resolution + dedup) — do not re-implement.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Fail loud, gated on migration.&lt;/strong&gt; Turn silent drops into warnings/errors; only flip fail-&lt;em&gt;closed&lt;/em&gt; once the population is 100% migrated (fragment-cascade readiness, &lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt; backfill completeness).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Enforce with parity/totality gates.&lt;/strong&gt; A gate that binds loader↔resolver recursion &lt;em&gt;and&lt;/em&gt; covers &lt;code dir=&quot;auto&quot;&gt;str&lt;/code&gt;-keyed maps (which escape today’s &lt;code dir=&quot;auto&quot;&gt;test_kind_mapping_totality.py&lt;/code&gt;), so fixes cannot re-drift.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;4-mission-decomposition-a-program-not-one-mission&quot;&gt;4. Mission decomposition (a program, not one mission)&lt;/h2&gt;
&lt;blockquote&gt;
&lt;p&gt;This is genuinely a &lt;strong&gt;program&lt;/strong&gt; rolling up to #3530/#3410. Below is the recommended cut. The two enabling missions can start immediately in parallel; the completeness mission lands last because it re-ledgers golden counts.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;h3 id=&quot;enabling-layer--land-first-in-parallel&quot;&gt;Enabling layer — land first, in parallel&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;M1 — Single-authority resolution parity&lt;/strong&gt; · covers &lt;strong&gt;#3490, #3426, #2981&lt;/strong&gt; · effort &lt;strong&gt;M&lt;/strong&gt;
One recursion authority read by both loader and resolver (make org/project discovery recursive to match built-in; delete the redundant Styleguide/Asset overrides; fix &lt;code dir=&quot;auto&quot;&gt;agent_profiles/repository.py&lt;/code&gt;’s third divergence site). One derived plural↔singular kind map from &lt;code dir=&quot;auto&quot;&gt;ArtifactKind&lt;/code&gt;, collapsing the 4 charter duplicators (&lt;code dir=&quot;auto&quot;&gt;activations.py:178/194&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;_activation_render.py:271/112&lt;/code&gt;). Ship the shared &lt;strong&gt;parity/totality gate&lt;/strong&gt; that (a) binds {kinds the loader scans recursively} == {kinds the resolver returns &lt;code dir=&quot;auto&quot;&gt;recursive=True&lt;/code&gt;}, and (b) covers &lt;code dir=&quot;auto&quot;&gt;str&lt;/code&gt;-keyed maps. Self-contained; &lt;strong&gt;no golden-count ripple&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;M2 — DRG read-path bridge&lt;/strong&gt; · covers &lt;strong&gt;#3572, #3573&lt;/strong&gt; · effort &lt;strong&gt;M&lt;/strong&gt;
Route &lt;code dir=&quot;auto&quot;&gt;load_validated_graph&lt;/code&gt;’s org layer through &lt;code dir=&quot;auto&quot;&gt;merge_three_layers(built_in, org_fragments, project)&lt;/code&gt; (add an &lt;code dir=&quot;auto&quot;&gt;org_fragments&lt;/code&gt; param populated by &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; via &lt;code dir=&quot;auto&quot;&gt;charter.drg.load_org_drg&lt;/code&gt;; build-time callers stay inert). Re-scope the D-005 warning to fire only when a root ships &lt;em&gt;neither&lt;/em&gt; a root graph &lt;em&gt;nor&lt;/em&gt; a fragment. Flip &lt;code dir=&quot;auto&quot;&gt;TestGraphlessPackWithFragmentEdgeIsInvisibleToCascade&lt;/code&gt; to assert the edge &lt;strong&gt;does&lt;/strong&gt; cascade, and &lt;strong&gt;atomically&lt;/strong&gt; update the #3573 validator so its message never becomes a lie. Enabling for all org-authored reach.&lt;/p&gt;
&lt;h3 id=&quot;reachdelivery-layer&quot;&gt;Reach/delivery layer&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;M3 — Operating-procedures: validate → triage → data-drive&lt;/strong&gt; · covers &lt;strong&gt;#2994, #3352, #3488(op-proc channel)&lt;/strong&gt; · effort &lt;strong&gt;L&lt;/strong&gt; · &lt;em&gt;hard internal order&lt;/em&gt;
(1) Load-time validator: every &lt;code dir=&quot;auto&quot;&gt;operating-procedures&lt;/code&gt; entry must resolve to a real node (converts 36 silent fictional refs to loud). (2) Triage the 36 (author/repoint/delete). (3) Data-drive the extractor to emit &lt;code dir=&quot;auto&quot;&gt;agent_profile --requires--&gt; procedure&lt;/code&gt;, guarded to resolvable targets, retiring the hand-pinned &lt;code dir=&quot;auto&quot;&gt;_CURATED_ARTIFACT_EDGES&lt;/code&gt;. &lt;strong&gt;Must not&lt;/strong&gt; emit before targets resolve (else 36 dangling edges → &lt;code dir=&quot;auto&quot;&gt;assert_valid&lt;/code&gt; failure).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;M4 — Deliver loaded doctrine to the dispatched agent&lt;/strong&gt; · covers &lt;strong&gt;#3489, #3176, #3389, #3488(render half)&lt;/strong&gt; · effort &lt;strong&gt;M&lt;/strong&gt;
Fix the delivery-table/render no-ops: give &lt;code dir=&quot;auto&quot;&gt;GLOSSARY_PACK&lt;/code&gt; a real slot + term-list render row (or ratify document-only) and restore the “stated reason” for every remaining &lt;code dir=&quot;auto&quot;&gt;None&lt;/code&gt; row; render procedure step &lt;code dir=&quot;auto&quot;&gt;description&lt;/code&gt; (not just &lt;code dir=&quot;auto&quot;&gt;title&lt;/code&gt;); decide styleguide/toolguide inline-body-vs-pointer. Add the &lt;code dir=&quot;auto&quot;&gt;#3176&lt;/code&gt; builder-overlay seam so &lt;code dir=&quot;auto&quot;&gt;.kittify/agent_profiles&lt;/code&gt; is reachable. Promote &lt;code dir=&quot;auto&quot;&gt;procedures[]&lt;/code&gt; to a typed array in &lt;code dir=&quot;auto&quot;&gt;context --json&lt;/code&gt; (&lt;strong&gt;versioned-contract bump&lt;/strong&gt; — do deliberately). Org-authored glossary/op-proc acceptance is &lt;strong&gt;gated on M2&lt;/strong&gt;.&lt;/p&gt;
&lt;h3 id=&quot;completeness-layer--land-last&quot;&gt;Completeness layer — land last&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;M5 — Kind-complete cascade + orphan wiring&lt;/strong&gt; · covers &lt;strong&gt;#2829, #3009 residual&lt;/strong&gt; · effort &lt;strong&gt;L&lt;/strong&gt; · &lt;em&gt;ADR-worthy&lt;/em&gt;
Expand the cascade relation set to follow the action-hop relations (&lt;code dir=&quot;auto&quot;&gt;scope&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;instantiates&lt;/code&gt;; keep &lt;code dir=&quot;auto&quot;&gt;in_tension_with&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;rejects&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;delegates_to&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;applies&lt;/code&gt; excluded), so activating a mission_type reaches its dependent kinds. Author real inbound edges for the ~5 remaining &lt;code dir=&quot;auto&quot;&gt;_ACTIVATED_BUT_ORPHANED&lt;/code&gt; artefacts (or mark direct-activation-only). Lands &lt;strong&gt;last&lt;/strong&gt; so it re-ledgers golden counts &lt;strong&gt;once&lt;/strong&gt;, atop M2’s fragment edges.&lt;/p&gt;
&lt;h3 id=&quot;carve-out&quot;&gt;Carve-out&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;M6 — Project-tier DRG node emission&lt;/strong&gt; · covers &lt;strong&gt;#3038&lt;/strong&gt; · effort &lt;strong&gt;L&lt;/strong&gt;
Emit hand-authored project-tier &lt;code dir=&quot;auto&quot;&gt;agent_profile&lt;/code&gt; artefacts as DRG nodes (a filesystem-walk emitter, not just extending the answer-driven map). Reuses M1’s single-authority discipline. &lt;strong&gt;Asset half deferred behind #3037&lt;/strong&gt; (asset has no resolution/install path). Convert &lt;code dir=&quot;auto&quot;&gt;project_drg._KIND_TO_NODE_KIND&lt;/code&gt; to enum-keyed so the totality gate covers it.&lt;/p&gt;
&lt;h3 id=&quot;cleanup-rider-rideable-on-any-mission&quot;&gt;Cleanup rider (rideable on any mission)&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;#3575&lt;/strong&gt; (delegation swap &lt;code dir=&quot;auto&quot;&gt;builtin_missions_root → pack_paths.built_in_missions_root&lt;/code&gt;, S) and the &lt;strong&gt;doc-reference half of #3574&lt;/strong&gt; (replace the in-code deferral with an issue link, S). The &lt;strong&gt;#3574 resolver cutover&lt;/strong&gt; (fail-loud &lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt; resolution, &lt;strong&gt;L&lt;/strong&gt;, gated on backfill completeness) belongs to its own later pack-identity mission — not this program.&lt;/p&gt;
&lt;h2 id=&quot;5-sequencing&quot;&gt;5. Sequencing&lt;/h2&gt;
&lt;div class=&quot;expressive-code&quot;&gt;&lt;figure class=&quot;frame not-content&quot;&gt;&lt;figcaption class=&quot;header&quot;&gt;&lt;/figcaption&gt;&lt;pre data-language=&quot;plaintext&quot;&gt;&lt;code&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;M1 ─┐                         (enabling, no golden ripple)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;M2 ─┼─► M3 ─► M4(org reach)    (M3/M4 org acceptance gated on M2; M3 before its data-drive)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;    &lt;/span&gt;&lt;/span&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;└─► M4(render/builder) ────►&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;                          &lt;/span&gt;&lt;/span&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;M5    (LAST — re-ledger once, atop M2)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#d6deeb;--1:#403f53&quot;&gt;M1 ─► M6                        (reuses M1 discipline)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/figure&gt;&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;M1 ∥ M2&lt;/strong&gt; first — both enabling, independent.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;M3&lt;/strong&gt; needs the validated node universe; its data-drive step must follow triage.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;M4&lt;/strong&gt; render/builder parts are independent; its glossary/op-proc &lt;em&gt;org&lt;/em&gt; acceptance waits on M2.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;M5&lt;/strong&gt; is the only golden-count-heavy change on cascade reach — land it last so counts move once (with M2’s edges already in).&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;6-scope&quot;&gt;6. Scope&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;In:&lt;/strong&gt; charter resolution / cascade / delivery; org &amp;#x26; project doctrine loading (recursion, tier node-admission); validation honesty &amp;#x26; canonical-vocab consistency.
&lt;strong&gt;Out:&lt;/strong&gt; asset resolution/install (&lt;strong&gt;#3037&lt;/strong&gt;); the manifest-reliability boundary (&lt;strong&gt;#3388/#3412&lt;/strong&gt; — #3530 explicitly marks these adjacent, different root cause); the &lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt; fail-loud &lt;em&gt;flip&lt;/em&gt; (needs a separate backfill-completeness gate).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Already landed — do not redo:&lt;/strong&gt; most of &lt;strong&gt;#3009&lt;/strong&gt; (membership set replacing the bare count, wire-8-delete-1, reachability companion); the service/repository half of &lt;strong&gt;#3038&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;PROJECT_KIND_DIRS&lt;/code&gt;); &lt;strong&gt;#2981&lt;/strong&gt;’s cited WP08 precedent; &lt;strong&gt;#3576&lt;/strong&gt; fixes only the #3575 &lt;em&gt;comment&lt;/em&gt;, not the delegation.&lt;/p&gt;
&lt;h2 id=&quot;7-risks&quot;&gt;7. Risks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Unconditional recursion&lt;/strong&gt; loads nested YAML not intended for loading — mitigated because globs are kind-specific (&lt;code dir=&quot;auto&quot;&gt;*.tactic.yaml&lt;/code&gt;); verify &lt;code dir=&quot;auto&quot;&gt;.provenance/*.yaml&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;.md&lt;/code&gt; files are not captured.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Cascade relation-set expansion (M5) ripples golden counts&lt;/strong&gt; — sequence adjacent to M2, re-ledger once.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Widening styleguide/toolguide bodies blows the NFR-001 token budget&lt;/strong&gt; — the reason they were pointer-only; treat as a doctrine decision, not a silent widen.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Data-driving op-procedures before triage&lt;/strong&gt; mints 36 dangling edges → &lt;code dir=&quot;auto&quot;&gt;assert_valid&lt;/code&gt; failure. Order is load-bearing.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;str&lt;/code&gt;-keyed maps escape the totality gate&lt;/strong&gt; — re-key on &lt;code dir=&quot;auto&quot;&gt;ArtifactKind&lt;/code&gt; or add per-universe bidirectional equality tests.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Premature &lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt; fail-closed&lt;/strong&gt; strands un-backfilled packs (why the flip is out of scope here).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Watch the baseline-red gotcha&lt;/strong&gt; — run new tests on the merge-base to confirm green-before.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;8-open-questions--operator-decisions-at-spec-time&quot;&gt;8. Open questions — operator decisions at spec time&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Recursion (M1):&lt;/strong&gt; make org/project discovery unconditionally recursive (simplest, matches built-in) or keep a per-kind flag shared by loader+resolver?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;anti_pattern&lt;/code&gt; (M1):&lt;/strong&gt; is the charter-activatable kind map &lt;strong&gt;9&lt;/strong&gt; kinds (faithful to &lt;code dir=&quot;auto&quot;&gt;CHARTER_KIND_TOKENS&lt;/code&gt;) or &lt;strong&gt;10&lt;/strong&gt; (preserving today’s &lt;code dir=&quot;auto&quot;&gt;activations.py&lt;/code&gt; &lt;code dir=&quot;auto&quot;&gt;anti_patterns&lt;/code&gt; entry)? — the #2976 ruling; must not be silently pre-empted.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;--include&lt;/code&gt; selector vocab (M1):&lt;/strong&gt; widen the selector surface to accept &lt;code dir=&quot;auto&quot;&gt;glossary_pack&lt;/code&gt;/&lt;code dir=&quot;auto&quot;&gt;anti_pattern&lt;/code&gt;, or accept a correctly-derived-but-unrunnable stanza?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;#3573 (M2):&lt;/strong&gt; fold into M2 for an atomic flip (&lt;strong&gt;recommended&lt;/strong&gt;) or ship standalone with removal gated on #3572?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Glossary delivery (M4):&lt;/strong&gt; real delivery slot vs document-only exclusion? inline terms vs surface-list + fetch pointer (token budget)?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;styleguide/toolguide (M4):&lt;/strong&gt; grant a budgeted inline body, or ratify pointer-only and make it discoverable in schema/docs?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;operating-procedures field (M3):&lt;/strong&gt; wire it into the procedure channel, or deprecate/rename? (either way, add the dead-entry diagnostic.)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;#3038 (M6):&lt;/strong&gt; filesystem-walk emitter vs synthesis-answer-driven node presence? asset explicitly behind #3037?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;#3389 asset asymmetry (M4):&lt;/strong&gt; after &lt;code dir=&quot;auto&quot;&gt;procedures[]&lt;/code&gt;, is &lt;code dir=&quot;auto&quot;&gt;asset&lt;/code&gt; deliberately reference-only forever, or a follow-up sixth typed array? — state it in the contract.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code dir=&quot;auto&quot;&gt;pack_id&lt;/code&gt; fail-loud arming (later):&lt;/strong&gt; what backfill-completeness gate + doctor/audit surface proves 100% coverage before the flip?&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;9-recommended-first-mission&quot;&gt;9. Recommended first mission&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Lead with M1 (Single-authority resolution parity), M2 in parallel.&lt;/strong&gt; M1 is self-contained, ripples no golden counts, closes the highest-impact silent-loss defect (#3490’s 71% tactic undercount + #3426), and — most importantly — installs the &lt;strong&gt;parity/totality gate&lt;/strong&gt; that protects every later mission from re-drift. M2 is the other enabling fix and unblocks all org-authored reach, so start it alongside. Everything else sequences behind these two.&lt;/p&gt;</content>
</entry>
