Mission-Type Completion Arc — Pre-Spec Research

Scope. Research supporting three sibling missions under epic #2652, produced by a 3-lens pre-spec squad (architect-alphonso, python-pedro, paula-patterns) verified against main HEAD e162a8900 (post-#2676 merge). Owned as planning deliverables by the two missions specced from it: mission-type-drg-edges-01KXKY2N (#2677) and activation-driven-enumeration-01KXKY7J (#2657 + #2659-E1). Transient by intent — fold into the owning missions' kitty-specs/ artifacts on close.

1. The governing ADR maps the whole arc to sequenced slices

ADR 2026-07-15-1 ("Doctrine Offers, Charter Activates, Runtime Consumes Only Activated") states the binding rule and its slices. Our three issues map onto them:

ADR slice Meaning Issue Mission
S0 mission_type (+ mission_step_contract) as first-class DRG nodes — and their edges #2651 minted nodes; #2677 wires edges A (drg-edges)
S1 init provisions the default charter (packs/default.yaml) #2657 B (enumeration)
S2 collapse the read-time "absent key → all" fallback onto default.yaml #2657 B
enum canonicalise enumeration onto the activation set (E1) #2659-E1 B
S3 gate the runtime _runtime_template_key/get_or_start_run path (E2) #2659-E2 C (deferred)
S4 wire the template_set slot (templates as activation-scoped config) later future

Guards to land with the work (ADR G1–G6): G2 (built-in list derived from MissionTypeRepository.default()

  • drift test) was already delivered by #2676 for show_origin.py/home.py. G1 (single availability seam), G3 (default charter one artefact) are Mission B's. G4/G5/G6 are downstream.

2. Brief A — DRG-edge design (#2677 / Mission A)

Bottom line: one edge class resolves all 8 orphans with zero new node kinds.

  • The 8 orphans are mission_type:{software-dev,documentation,research,plan} + action:plan/{plan,research,review,specify}, minted edge-less by #2651 (src/doctrine/drg/models.py:46 — "nodes only, no edges yet"). They push the shipped-graph orphan count to 18 > ceiling 14 (test_shipped_graph_orphan_count_within_documented_residual, red on main).
  • Minimal-correct fix: emit mission_type:X → action:X/<step> with relation requires, sourced from each type's action_sequence (mission_types/<id>.yaml), targeting action:* nodes that already exist (24 of them). mission_type:plan → action:plan/{specify,research,plan,review} de-orphans both the plan type node (outbound) and the 4 plan action nodes (inbound). Count 18 → 10 ≤ 14 — do not raise the ceiling.
  • requires is correct: composition-of-mandatory-ordered-parts, the relation the charter cascade traverses (making the currently-no-op mission-type cascade meaningful); no cycle (actions only emit scope outbound); validator-clean (all sequence steps have index dirs → no dangling target).
  • The other operator-intended targets need node populations that don't exist: mission_step_contract (0 nodes; but steps = action nodes already), WP template: nodes (16 exist but wrong population; template_set null for 3/4 types — #883 template deferral), asset (0 nodes — assets/built-in/ is just a README), guard/gate (no NodeKind at all). All honest future scope; none touch the orphans.
  • Generator workflow: add extract_mission_type_edges beside _discover_mission_type_nodes (extractor.py:768), emit into all_edges before calibrate/sort; spec-kitty doctrine regenerate-graph rewrites + commit graph.yaml (byte-identity freshness gate). Re-pin the S0 placeholder test_mission_type_nodes_have_no_incident_edges (invert), update the models.py:46 comment, note the residual doc. Open forks: relation requires vs instantiates (recommend requires); retrospect actions (in no action_sequence) — leave (already non-orphan).

3. Brief B — Default-charter + enumeration current state (#2657 + #2659 / Mission B)

  • #2657 real remaining scope (smaller than the epic implies): the dead except ImportError → CANONICAL branch is already gone (retired by #2669); the behavioral "absent key → all four" fallback remains, single-sourced at pack_context.py:277. packs/default.yaml, load_default_pack_activation_ids(), and the legacy-provisioning migrations (m_3_2_0rc35_default_charter_pack, ..._activate_builtin_mission_types) all exist. Remaining = S2 (read-time authority → default-pack data; delete _BUILTIN_* as a fallback source; keep fail-closed) + S1 (fresh-init provisioning — the real gap; no init path writes the keys today).
  • #2659 splits E1/E2: E1 (route list_available_missions mission.py:489 + discover_missions mission.py:806 + _packaged_missions_dir through existing_mission_types) is low-risk and needs only #2657. E2 (flip _build_discovery_context runtime_bridge_io.py:230-241 root specify_cli→doctrine + activation-gate) is hot-loop and blocked additionally on #2658 (templates-as-config). The doctrine plan sidecar even lacks the top-level steps: block the runtime schema needs — a disposable doctrine-vs-specify_cli parity scaffold is required before the flip.
  • Ordering: #2657 MUST precede E1 — the E1 filter is untestable while an implicit absent-key fallback re-expands to all four (masks deactivation).

4. Brief C — Mission decomposition (3 tracks, not one)

The three issues touch three subsystems with no shared write-seam (doctrine/drg vs charter vs specify_cli/runtime). Recommended decomposition:

  • Mission A = #2677 — standalone doctrine/drg; clears the red main gate; ship first.
  • Mission B = #2657 + #2659-E1 — charter + specify_cli; internally #2657(S2→S1) → E1.
  • Mission C = #2659-E2 — deferred, gated on #2658 + parity scaffold.

Foldable-issue map: #2658 (prerequisite for E2 only — do not fold), #2660/#2661 (downstream/terminal), #1923 (coordinate only on the DOCUMENTED_ORPHAN_RESIDUAL = 14 constant — #2677 wires its 8 orphans, #1923 curates the other 14). Post-#2676 drift: #2657's cited mission_type_profiles.py:388 is now the existing_mission_types docstring; the fallback moved to pack_context.py:277.

Edit this page