Mission Specification: Dead-Port Disposition

Mission Branch: dead-port-disposition-01M1TZVN Created: 2026-09-06 Status: Draft (proto-mission — sliced at spec level; no plan.md or tasks yet; open decisions carried explicitly below) Input: User description: "Retire the dormant mission-DSL and its hot-path transitions dependency, repair the glossary-runner design story, and stage the ADR-gated RuntimeEventEmitter cleanup — adjudicating and cleaning the core loop's dead and dormant hook seams."

Ground truth: research/29-missionB-research-dossier.md (Researcher Robbie, verified at HEAD e721763759). Where this spec and the dossier disagree, the dossier wins; flag the drift.

Sequencing pins (binding): 1. This mission runs AFTER Mission A fsm-write-path-integrity-01M1TZV6 — the four shared runtime files (runtime_bridge_io.py, _internal_runtime/engine.py, decision.py, runtime_bridge_engine.py) are NOT parallel-safe; no edits to them until Mission A's WPs on them are merged. Still open and unaffected by this landing pass. 2. SATISFIED as of 2026-09-06. The test_layer_rules.py residue item (FR-013) was gated on PR #3888 (which modifies test_layer_rules.py and test_runtime_charter_doctrine_boundary.py); #3888 merged 2026-09-06. FR-013/C-002 may proceed. 3. SATISFIED as of 2026-09-06. The RuntimeEventEmitter disposition ADR was an INPUT DEPENDENCY on PR #3898 — docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md. #3898 merged 2026-09-06 and the ADR now carries status: Accepted. The WP03 gate is OPEN: FR-012/US4 execute per the ADR's chosen option (rewire-ready consolidation, which deletes event_emitter.py) rather than shipping the shrunk residue-only variant.

Ground-Truth Errata

This mission's citations were re-verified against main at 3b38073a32 during the 2026-09-07 landing pass (PR #3904). The dossier (research/29-missionB-research-dossier.md) carries the same drifted or imprecise values for every row below and is left unmodified — it is an immutable evidence snapshot anchored at e721763759 — but for the entries in this table, the spec (as corrected here) overrides the dossier, notwithstanding the general "where this spec and the dossier disagree, the dossier wins" rule.

Dossier/spec saidActually true (verified 3b38073a32)Why it changed
pyproject.toml:82 is the transitions>=0.9.2 entry (throughout the spec)transitions>=0.9.2 is now at line 81; line 82 is jsonschema>=4.0 — mission-config validation's dependency, not part of this retirement. Every citation now names the entry, not a line number, so it cannot rot againDrifted since e721763759; PR #3899 (merged 2026-09-06) removed truststore>=0.10.4 above it and shifted the block
test_layer_rules.py:65-73 is the constitution exclusion (US3, FR-013)The _EXCLUDED_FROM_LAYER_ENFORCEMENT frozenset holding the constitution entry now spans :66-74Drifted since e721763759
test_no_dead_symbols.py:718-727 holds the three dead-symbol pins (US1-AS5, C-005, SC-006)The three pins are at :719-727; line :718 closes the unrelated clear_mission_brief pin — deleting it would leave the SymbolKey( opened at :716 unterminated (SyntaxError)Drifted since e721763759
"engine ×16" names the engine in the 29-occurrence sync_emitter map (US4)_internal_runtime/engine.py has zero sync_emitter sites; the 16 are in runtime_bridge_engine.py (the 13 in runtime_bridge.py were already named correctly)Wrong from the start; ambiguous engine reference
glossary_hook.py:16-17 is "contradicted eleven lines above its own bootstrap" (US2)The bootstrap is at :127, which is 111 lines below :16, not elevenWrong from the start
Sequencing pins 2 and 3 (header block): PR #3888 open; PR #3898 open with the ADR status: Proposed#3888 merged 2026-09-06; #3898 merged 2026-09-06 and the ADR now carries status: Accepted — both gates are OPEN (see Sequencing pins above, FR-012, C-003, Open Decision 7)Resolved after the dossier's e721763759 snapshot

User Scenarios & Testing (mandatory)

User Story 1 - spec-kitty next stops paying an import tax for a DSL nothing uses (Priority: P1)

As a maintainer, when any agent or operator runs spec-kitty next, the runtime should load only what the live loop needs. Today specify_cli/mission_v1/__init__.py:20-28 (compat at :20) eagerly imports all five submodules (compat, events, guards, runner, schema), so the live observability import — runtime/next/next_invocation_lifecycle.py:332 (emit_event) — plus the dormant read_events site (runtime/next/decision.py:200, inside the zero-caller derive_mission_state this mission's FR-015 slates for deletion) drags the transitions library (plus six) into sys.modules on every invocation, in service of a state-machine DSL with zero production callers (~1,190 test-only LOC). This story retires the mission-DSL v1 runtime AND breaks the eager import chain — the dossier's proven constraint is "both, together, or neither": deleting files without trimming __init__ frees nothing; trimming __init__ without deleting the files leaves the dead code in place.

Why this priority: It is the mission's bulk and the only slice that removes a third-party dependency from the hot path of every spec-kitty next invocation. It is unconditional (no ADR gate) and independently shippable.

Independent Test: In a subprocess-isolated test, import specify_cli.mission_v1.events completes with transitions and six absent from sys.modules, while emit_event/read_events still function for their two live consumers; transitions is absent from pyproject.toml and uv.lock; clean-install CI is green.

Acceptance Scenarios:

1. Given the current tree (eager block intact), When the red-first import-hygiene test runs import specify_cli.mission_v1.events in an isolated subprocess and inspects sys.modules, Then it FAILS (red) — proving the test observes the defect before any change lands. 2. Given the retirement has landed, When the same test runs, Then transitions and six are absent from sys.modules and the test is green — and it remains in the suite as a permanent ratchet against re-eager-ing (an enduring negative invariant, not a disposable parity test). 3. Given the retirement has landed, When spec-kitty next is invoked on a live mission, Then the MissionNextInvoked observability write (emit_event) and the deprecated derive_mission_state reader (read_events) behave exactly as before — events.py survives as live code. 4. Given the transitions>=0.9.2 entry in [project].dependencies (pyproject.toml) is deleted and uv lock regenerated (never hand-edited), When the clean-install-verification job and tests/architectural/test_pyproject_shape.py run, Then both pass; six remains in the lock via its python-dateutil edge. 5. Given the settled retirement extent (MIN or FULL — see Open Decisions), When the deletion PR lands, Then the same PR removes the three dead-symbol pins at tests/architectural/test_no_dead_symbols.py:719-727, updates the _gate_coverage.py:1456 mission_v1 shard row if tests/specify_cli/mission_v1/ empties, and (Option FULL only) removes the ("specify_cli.mission_v1.guards", "read_wp_frontmatter") entry from the fold gate at tests/specify_cli/test_wp_frontmatter_fold.py:60 — no gate goes red or vacuous. 6. Given tests/missions/test_mission_v1_events_unit.py imports runner at :228-304, When the blast-radius tests are reworked, Then that file is trimmed, not deleted — its events half covers live code.

User Story 2 - A reader of kernel/glossary_runner.py is told the truth about who registers (Priority: P2)

As a maintainer reading the kernel glossary-runner seam, I should find documentation that matches the mechanism that actually runs. The registry is LIVE (this is a repair, NOT a deletion): the only production register()/get_runner() path is the lazy self-bootstrap inside charter/offering/missions/glossary_hook.py:127-137 (get_runner() → on None, import_module("glossary.attachment") → register(GlossaryAwarePrimitiveRunner) → retry). Yet four verified sites tell three variants of a fiction — that specify_cli (or glossary) registers the runner at import/startup time: kernel/glossary_runner.py:14-16 (+ the stale provider block :33-38 and pre-relocation dependency diagram), charter/offering/missions/glossary_hook.py:16-17 (contradicted 111 lines above its own bootstrap at :127), kernel/README.md:18, and kernel/__init__.py:38-42. Separately, the hook's FR-020 claim ("glossary_check defaults to enabled") is enforced nowhere in the live mission loop: execute_with_glossary has zero production call sites and zero glossary_check references exist in packs/ — the honesty note must say so.

Why this priority: The doctrine-layer hook self-bootstrap keeps working either way, but the documented fiction is exactly the rot shape (green tests over an undocumented-in-truth mechanism) that produced the dead-port confusion this mission cleans up. Cheap (docs-shaped, ~0.5 day), independent of every gate.

Independent Test: Read the four sites — each describes the self-bootstrap (or the settled alternative, see Open Decisions); run the new behavior test — after execute_with_glossary executes with no prior registration, get_runner() returns the concrete runner.

Acceptance Scenarios:

1. Given the four fiction sites at their verified locations, When the repair lands, Then each documents the real contract per the settled Open Decision — recommendation carried from report 27 (B-WP02): option (a), canonicalize the self-bootstrap — and the "graceful degradation when no runner is registered" story is restated as "degradation only when glossary.attachment itself is unimportable". 2. Given a fresh process with no prior registration, When execute_with_glossary runs, Then get_runner() returns the concrete GlossaryAwarePrimitiveRunner — pinned as a behavior test extending tests/doctrine/missions/test_glossary_hook.py, turning today's incidental mechanism into a pinned invariant. 3. Given the FR-020 enforcement gap (zero production call sites for execute_with_glossary; zero glossary_check in packs/), When the repair lands, Then an honest record exists in both a code-adjacent note (e.g. a .. note:: in the hook) and a tracker note on the glossary workstream (#1868) — silently keeping green tests over a dead production path is explicitly rejected. 4. Given the optional three-link re-export chain tidy (charter/offering/missions/__init__.py:10 → charter/primitives.py:14 → specify_cli/missions/__init__.py:28), When it would require touching the escalated live-collision pins in test_no_dead_symbols.py, Then it is skipped — it is optional, never load-bearing.

User Story 3 - Gate-adjacent residue is routed and retired without pin-hash races (Priority: P3)

As a maintainer, the small confirmed-dead residue that couples to this mission's gate files gets cleaned here (once its external gates open), and residue with zero coupling goes to the quick-wins PR — one owner per gate file, no two PRs racing on test_no_dead_symbols.py hash pins. The quick-wins PR is PR #3899 (draft): it covers the scripts/dep/symlink/prose set (CODEOWNERS, CONTRIBUTING symlink, the orphan scripts + their test, truststore in pyproject+uv.lock, Makefile/pytest.ini/testing-parallel prose) — the src/mission_runtime residue is owned HERE. In scope here: the stale constitution exclusion at test_layer_rules.py:66-74 (the _EXCLUDED_FROM_LAYER_ENFORCEMENT frozenset; a package that no longer exists — mission 063 done), sequenced strictly after PR #3888; the 13 un-demoted test-only __all__ exports (demotion from package __all__, never deletion — content_present_at_primary_tip and get_packs_root_default have in-package src callers); the team_projection/ tombstone; and the ActionContext alias (two files: mission_runtime/context.py:~338 AND the __getattr__ block in mission_runtime/__init__.py:127,139-142) — all re-routed into this mission's residue slice, since #3899's verified file list does not carry them. Routed to #3899 (not here): truststore and the rest of its scripts/dep/symlink/prose scope.

Why this priority: Low-risk hygiene that prevents vacuous gates, but valueless if it races another PR on the same gate file; sequencing discipline is the whole story.

Independent Test: After #3888 merges and this slice lands, test_layer_rules.py carries no exclusion for a nonexistent package; every residue item is traceably placed (this mission, the quick-wins PR, or explicitly deferred) with its gate re-pinned in the same PR.

Acceptance Scenarios:

1. Given PR #3888 has merged, When the constitution exclusion at test_layer_rules.py:66-74 is deleted, Then the layer-rules gate stays green and the rebase over #3888's _RUNTIME_ALLOWED_SPECIFY_CLI ledger changes is trivial. 2. Given the residue-placement decision routes the 13 __all__ demotions into this mission, When they land, Then they ride the same PR as this mission's other test_no_dead_symbols.py edits, with matching re-pins — never a second concurrent PR on that gate file. 3. Given PR #3888 has NOT merged by the time this slice is ready, When the mission otherwise completes, Then the test_layer_rules.py hunk is held back (or minted as a follow-up), not landed first.

User Story 4 - The emitter seam is cleaned exactly as far as the ADR permits — and no further (Priority: P4, ADR RESOLVED — PR #3898 merged and Accepted 2026-09-06; executes per the ADR)

As a maintainer, the RuntimeEventEmitter confusion — two classes with the same name (runtime/next/event_emitter.py:23, a concrete 88-LOC no-op, vs runtime/next/_internal_runtime/events.py:67, the live Protocol with NullEmitter) — is resolved per the disposition ADR — PR #3898, docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md, currently status: Proposed. This mission's unconditional contribution is the verified record (safe under ANY disposition): the name-collision map; the 29-occurrence sync_emitter map (runtime_bridge_engine.py ×16, all annotations under a TYPE_CHECKING-only import at :80; runtime_bridge.py ×13, the sole runtime coupling — import :195, for_feature :1552/:2739, seed_from_snapshot :1614/:2745); and the live buffer flush-target defect (buffer swap runtime_bridge.py:2149-2150, flush into the inner no-op ctx.sync_emitter at :2187 instead of the DecisionGitLog-wrapped emitter_for_engine built at :1560-1565 — recorded, NOT touched). Execution — retire/rewire/merge, any rename, sync_emitter signature changes, the flush-target adjudication, the parity-oracle update at tests/runtime/test_bridge_parity.py:1131-1152 — proceeds now that the ADR (PR #3898) has merged AND been Accepted (2026-09-06, satisfying the WP03 gate; see Ground-Truth Errata). The ADR's chosen option is rewire-ready consolidation, which deletes event_emitter.py outright rather than correcting its docstring.

Why this priority: The disposition is owned elsewhere (input dependency); this mission must neither pre-empt it nor lose the verified facts that anchor it. Also gated behind Mission A (both emitter files are Mission-A-adjacent) — triple-gated, hence last.

Independent Test: The ADR landed (PR #3898 merged, Accepted 2026-09-06): its decision is executed with the parity oracle updated in the same PR and no name collision remains.

Acceptance Scenarios:

1. Given the ADR (PR #3898) is merged AND Accepted (satisfied 2026-09-06), When its decision is executed, Then the name collision is resolved per the ADR, the sync_emitter vocabulary/signature pass happens as ONE pass over the 29 sites (not two), the flush-target defect is adjudicated per the ADR (never silently frozen or silently fixed), and test_side_effect_sinks_are_actually_reached is updated in the same PR if the capture is removed. 2. Given any disposition, When work touches runtime_bridge_retrospective.py, Then _BufferingRuntimeEmitter (:69) is untouched — it is rollback machinery, the only thing preventing an unretractable MissionRunCompleted on a rolled-back terminal advance, and is not deletable under any disposition. 3. Given a removal that erases a subpackage's last live specify_cli edge, When the PR is assembled, Then the same PR carries the test_runtime_ledger_has_no_stale_entries ledger edit (and respects the widened doctrine-scan baseline pinning runtime_bridge_io.py/runtime_bridge_composition.py lazy reaches).

Edge Cases

  • ADR outcome is "explicit deferral": did not occur — the ADR (PR #3898) was Accepted 2026-09-06; this edge case is retained for historical completeness only (see Ground-Truth Errata; Open Decision 7 resolved).
  • The operator answers "yes, mission-DSL v1 is on a roadmap" (Open Decision 2): P1 converts from retirement to re-justification and the transitions pin stays — the mission's bulk re-scopes.
  • events.py relocation chosen (Open Decision 4): the move touches the two runtime lazy imports and the seam test tests/specify_cli/next/test_next_invocation_lifecycle_seam.py:49 — cheap now, costlier later; if kept in place, the package docstring must stop advertising the deleted DSL.
  • Option FULL chosen but the pack-DSL blocks are kept: the three mission.yaml states:/transitions: blocks lose their only interpreter (the schema-validation test halves are deleted); keeping them requires the in-file "uninterpreted" comment, and MISSION_COMPAT_IGNORED_FIELDS (mission.py:59-67,260-261) stays either way for third-party hybrid YAMLs.
  • Mission A's WP05 lane still open at tasks-time: the two dead DSL readers in decision.py (derive_mission_state:~187, evaluate_guards:~218 — zero callers; runtime_bridge_cores' evaluate_guards* family is an unrelated same-name function) are handed to that lane; otherwise a small follow-up owns decision.py alone after A merges (Open Decision 9).
  • Quick-wins PR and this mission both want the 13 __all__ demotions: whoever edits test_no_dead_symbols.py first takes them — the failure mode being prevented is two PRs racing on the same gate file's hash pins.
  • test_mission_v1_events_unit.py deleted wholesale by mistake: its :228-304 runner half dies with the DSL, but its events half covers live production code — trim, don't delete.

Requirements (mandatory)

Functional Requirements

IDTitleUser StoryPriorityStatus
FR-001Break the full eager import blockAs a maintainer, I want specify_cli/mission_v1/__init__.py:20-28 (compat at :20) reworked so that importing mission_v1.events loads neither transitions nor six nor the other four submodules — the whole eager block (all five submodules), not two lines.HighOpen
FR-002Retire the mission-DSL v1 runtimeAs a maintainer, I want runner.py, compat.py, and the load_mission/load_mission_by_name dispatch deleted (plus guards.py/schema.py under Option FULL — Open Decision 1) so no dead DSL code remains under the settled extent.HighOpen
FR-003Preserve the live events.py surfaceAs a runtime consumer, I want emit_event and read_events to keep working unchanged for next_invocation_lifecycle.py:332 and decision.py:200 so the retirement is invisible to the live loop. Note: after FR-015 deletes derive_mission_state, the sole live events.py consumer is emit_event at next_invocation_lifecycle.py:332; FR-003's preservation duty then narrows accordingly.HighOpen
FR-004Red-first import-hygiene ratchetAs a maintainer, I want a subprocess-isolated test — red today — asserting import specify_cli.mission_v1.events leaves transitions out of sys.modules, kept permanently as a ratchet against re-eager-ing.HighOpen
FR-005Drop the transitions dependencyAs a maintainer, I want the transitions>=0.9.2 entry in [project].dependencies (pyproject.toml) removed and uv.lock regenerated via uv lock so the dependency leaves the product (six stays via python-dateutil).HighOpen
FR-006Rework the DSL test blast radiusAs a maintainer, I want the extent-appropriate test files deleted or trimmed (per the dossier §1.3 table; test_mission_v1_events_unit.py trimmed, never deleted) so no test imports retired code.HighOpen
FR-007Pack-DSL honestyAs a pack author, I want the orphan states:/transitions: blocks in the three built-in mission.yamls either deleted or explicitly commented as uninterpreted (Open Decision 3), with MISSION_COMPAT_IGNORED_FIELDS retained, so the packs stop implying a live interpreter.MediumOpen
FR-008Repair the four glossary fiction sitesAs a reader of the kernel glossary seam, I want glossary_runner.py:14-16,33-38, glossary_hook.py:16-17, kernel/README.md:18, and kernel/__init__.py:38-42 rewritten to document the real registration mechanism (per Open Decision 5; recommendation: canonicalize the self-bootstrap).HighOpen
FR-009Pin the bootstrap contractAs a maintainer, I want a behavior test (extending test_glossary_hook.py) asserting that after execute_with_glossary runs with no prior registration, get_runner() returns the concrete runner — the documented contract becomes a pinned invariant.MediumOpen
FR-010FR-020 enforcement-honesty recordAs an operator, I want an honest record — in-code note plus a tracker note on #1868's glossary workstream — that glossary_check "enabled by default" is enforced nowhere in the live mission loop (zero production call sites; zero glossary_check in packs/).MediumOpen
FR-011Verified emitter record as ADR inputAs the ADR author, I want the name-collision map, the 29-occurrence sync_emitter map, the flush-target defect, and the spec_kitty_events 9.1.6 VOLATILE_EVENT_TYPES context preserved as verified inputs in a named home — the dossier §3, extracted as research/emitter-adr-inputs.md handed to the ADR author (PR #3898) — recorded, never executed ahead of the ADR.HighOpen
FR-012Execute the landed emitter ADRAs a maintainer, I want the ADR's disposition executed exactly (collision resolution, sync_emitter vocabulary as one pass, flush-target adjudication, parity-oracle update in the same PR) — the gate is satisfied: PR #3898 merged 2026-09-06 and the ADR (docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md) carries status: Accepted, so execution proceeds per the ADR's chosen option (rewire-ready consolidation) rather than the shrunk residue-only variant.MediumOpen
FR-013Delete the stale constitution exclusionAs a maintainer, I want test_layer_rules.py:66-74 (the _EXCLUDED_FROM_LAYER_ENFORCEMENT frozenset entry, excluding a package that no longer exists) removed — strictly after PR #3888 lands.LowOpen
FR-014Residue routing with single gate ownershipAs a maintainer, I want each residue item placed with single gate ownership: the quick-wins PR #3899 (draft) keeps its verified scripts/dep/symlink/prose scope, while the src/mission_runtime residue — the team_projection/ tombstone, the ActionContext alias (2 files), and the 13 __all__ demotions, none of which are in #3899's file list — rides this mission's gate-editing WP, so no two PRs race on test_no_dead_symbols.py hash pins (Open Decision 8).LowOpen
FR-015Dispose the decision.py dead DSL readersAs a maintainer, I want the two .. deprecated:: 2.0.0 functions (derive_mission_state, evaluate_guards) deleted by whichever lane owns the shared file post-Mission-A (Open Decision 9) — never edited in parallel with Mission A.LowOpen

Non-Functional Requirements

IDTitleRequirementCategoryPriorityStatus
NFR-001Hot-path import hygieneA subprocess running import specify_cli.mission_v1.events finishes with transitions and six absent from sys.modules — asserted by the permanent ratchet test on every CI run.PerformanceHighOpen
NFR-002Clean-install integrityAfter the dependency drop, the clean-install-verification CI job and tests/architectural/test_pyproject_shape.py pass with zero failures.ReliabilityHighOpen
NFR-003No vacuous or orphaned gatesEvery architectural gate edited by this mission (test_no_dead_symbols.py, _gate_coverage.py, test_wp_frontmatter_fold.py, test_layer_rules.py, the runtime ledger, the parity oracle) is updated in the same PR as the deletion it covers — zero red pins, zero pins asserting on deleted-and-forgotten surfaces.MaintainabilityHighOpen

Constraints

IDTitleConstraintCategoryPriorityStatus
C-001Mission-A sequencingNo edits to runtime_bridge_io.py, _internal_runtime/engine.py, decision.py, or runtime_bridge_engine.py until Mission A fsm-write-path-integrity-01M1TZV6's WPs on them are merged; the whole mission sequences after Mission A's shared-file lanes.TechnicalHighOpen
C-002PR #3888 sequencingThe test_layer_rules.py residue hunk (FR-013) lands only after PR #3888 merges; this mission rebases over it.TechnicalHighOpen
C-003Emitter ADR gate (satisfied)The RuntimeEventEmitter disposition ADR (PR #3898, docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md) merged 2026-09-06 and carries status: Accepted — the gate is OPEN. Emitter rewiring, deletion, renaming, and sync_emitter signature changes execute per the ADR's chosen option (rewire-ready consolidation, which deletes event_emitter.py); the flush-target defect is adjudicated per the ADR rather than left record-don't-touch.TechnicalHighOpen
C-004Dependency-removal mechanicsthe transitions>=0.9.2 entry in [project].dependencies (pyproject.toml) deleted; uv.lock regenerated with uv lock, never hand-edited; version bump in pyproject.toml + CHANGELOG.md entry (house rule); one dependency change per PR — truststore removal already landed via the quick-wins PR #3899 (merged 2026-09-06) and must not be re-bundled here.TechnicalHighOpen
C-005Same-PR gate editsThe DSL-retirement PR itself removes the three dead-symbol pins at test_no_dead_symbols.py:719-727, updates the _gate_coverage.py:1456 shard row if tests/specify_cli/mission_v1/ empties, and (Option FULL) the fold-gate entry at test_wp_frontmatter_fold.py:60.TechnicalHighOpen
C-006Rollback machinery is untouchable_BufferingRuntimeEmitter (runtime_bridge_retrospective.py:69) is not deletable under any disposition.TechnicalHighOpen
C-007Runtime-ledger disciplineAny removal erasing a subpackage's last live specify_cli edge carries the test_runtime_ledger_has_no_stale_entries ledger edit in the same PR; the widened doctrine-scan baseline's pins on runtime_bridge_io.py/runtime_bridge_composition.py lazy reaches are respected.TechnicalMediumOpen

Key Entities

  • mission-DSL v1 (src/specify_cli/mission_v1/): five submodules; only events.py (93 LOC) is production-live; runner.py/compat.py hold the only import transitions lines in the tree; ~1,190 test-only LOC total.
  • transitions dependency: the transitions>=0.9.2 entry in [project].dependencies (pyproject.toml), lock 0.9.3 with a six edge; loaded on every spec-kitty next via the eager chain.
  • Kernel glossary registry (kernel/glossary_runner.py + glossary_hook.py): LIVE via lazy self-bootstrap; its documented registration story is fiction at four sites; FR-020 enforcement is a documented-but-unwired claim.
  • RuntimeEventEmitter (×2): concrete no-op (event_emitter.py:23) vs live Protocol (_internal_runtime/events.py:67); disposition owned by the pending ADR.
  • Gate-adjacent residue: stale constitution exclusion; 13 test-only __all__ exports (demotion-only); team_projection/ tombstone and ActionContext alias (owned HERE — not in quick-wins PR #3899, whose scope is the scripts/dep/symlink/prose set incl. truststore).

Open Decisions (decision log — dossier §7; the specify interview is past — each decision carries its routing)

1. [NEEDS DECISION: DSL retirement extent — MIN vs FULL (§1.3) — BLOCKS-PLANNING; recorded working default: FULL (planning proceeds on it). Delete guards.py/schema.py too? Randy says FULL; Alphonso (report 24) keeps guards.py as shape-precedent — but the precedent is docstring prose that could cite git history instead. FULL forces the fold-gate edit (test_wp_frontmatter_fold.py:60) and kills the pack-DSL validator tests. Recommendation on the table: FULL, with gate_registry.py docstrings re-pointed at history.] 2. [NEEDS DECISION: Is mission-DSL v1 on any roadmap? — OPERATOR-ONLY, no default recorded. A "yes" converts the P1 slice from retirement to re-justification and keeps the transitions pin.] 3. [NEEDS DECISION: Pack DSL blocks (§1.4) — delete the orphan states:/transitions: blocks from the three mission.yamls, or keep-with-comment naming them uninterpreted? Does MISSION_COMPAT_IGNORED_FIELDS stay as third-party tolerance either way (recommended: yes)?] 4. [NEEDS DECISION: events.py future home — stays as specify_cli.mission_v1.events (a one-module package whose name outlives its DSL), or relocates (e.g. under status/mission_metadata)? Relocation touches the two runtime lazy imports and the seam test at test_next_invocation_lifecycle_seam.py:49 — cheap now, costlier later. If kept, does the package docstring get rewritten to stop advertising the deleted DSL?] 5. [NEEDS DECISION: Glossary contract choice (§2) — (a) canonicalize the self-bootstrap (docs-only; recommended by report 27) or (b) mint a real registration seam in specify_cli with the bootstrap as fallback (design change, no current driver)? And who owns the FR-020 honesty follow-through — a tracker note only, or does Mission C's design-codification arc absorb it?] 6. [NEEDS DECISION: Wire-or-retire for execute_with_glossary itself — OPERATOR-ONLY, no default recorded. Explicitly out of this mission (Non-goal 3), but the operator must confirm they accept "documented-but-unwired" as the steady state, else mint the feature issue now.] 7. [RESOLVED — ADR timing. The WP03 gate check ("ADR merged AND Accepted") is satisfied: PR #3898 merged 2026-09-06 and the ADR (docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md) carries status: Accepted. The P4 slice executes the ADR's chosen option (rewire-ready consolidation) rather than shipping the shrunk residue-only variant; no follow-up mission/WP is needed for the emitter execution itself.] 8. [NEEDS DECISION: Residue placement — the quick-wins PR is #3899 (draft), covering the scripts/dep/symlink/prose set; its verified file list carries NONE of the src/mission_runtime residue. Recorded default (planning proceeds on it): team_projection, ActionContext, AND the 13 __all__ demotions ride this mission's gate-editing WP — which also dissolves the test_no_dead_symbols.py hash-pin race, since #3899 edits no gate file. Operator may still re-route them onto an extended #3899.] 9. [NEEDS DECISION: decision.py dead-reader ownership — Mission A WP05 lane (if still open at tasks-time), or a small Mission B follow-up that owns the file alone after A merges?]

Non-Goals (verbatim from dossier §5)

1. No emitter rewiring, deletion, renaming, or sync_emitter signature changes before the disposition ADR lands. Mission B may carry the ADR's execution as a late WP only if the ADR has landed by tasks-time; otherwise that WP is minted separately. 2. No touches to Mission A's surfaces: status/emit.py, coordination/status_transition.py / transaction.py, status/aggregate.py, the writer census, dependency gating, lock work — and no parallel edits to the four shared files (runtime_bridge_io.py, _internal_runtime/engine.py, decision.py, runtime_bridge_engine.py) until Mission A's WPs on them are merged. 3. No wiring of execute_with_glossary into the live step executor (feature work, unowned). 4. No adjudication of _internal_runtime/{emitter,lifecycle,models}.py frozen re-exports (Category-6 contract renegotiation, shared-package-boundary owner). 5. No doctrine.py shim deletion (on schedule for 3.3.0 per shim-registry). 6. No truststore removal here — done via quick-wins PR #3899 (merged 2026-09-06); this mission must not re-touch it.

Success Criteria (mandatory)

Measurable Outcomes

  • SC-001: Importing specify_cli.mission_v1.events in an isolated subprocess leaves transitions (and six) out of sys.modules — the ratchet test is green and permanent in the suite.
  • SC-002: transitions is absent from pyproject.toml and uv.lock; six remains via python-dateutil; the clean-install-verification CI job and test_pyproject_shape.py are green.
  • SC-003: Zero import transitions (or from transitions import) lines remain anywhere in src/; the retired LOC matches the settled extent (~530 LOC MIN, ~1,190+ LOC FULL, per the §1.3 table); the two live consumers (next_invocation_lifecycle.py:332, decision.py:200) pass their existing tests unchanged.
  • SC-004: All four glossary fiction sites describe the settled registration contract; the bootstrap-contract behavior test is green; a grep across the four named files yields zero matches for register-at-import/startup phrasing (the primary, greppable check), with human review of the four sites as secondary.
  • SC-005: The FR-020 honesty record exists in both the code-adjacent note and the #1868 tracker note.
  • SC-006: The three dead-symbol pins (test_no_dead_symbols.py:719-727) are gone; every gate this mission touched is green with no vacuous entries; the operable race rule held — whoever edits test_no_dead_symbols.py first takes the demotions, and this mission checked open PRs against its gate files at PR-assembly time.
  • SC-007: The emitter slice executes the ADR's disposition — PR #3898 merged AND the ADR Accepted (2026-09-06): decision executed, collision gone, parity oracle updated in the same PR.
  • SC-008: The constitution exclusion is absent from test_layer_rules.py and landed after PR #3888 — never before.

Cross-References

  • #1868 — parent candidate epic for this mission (glossary-workstream tracker notes land there).
  • #2173 — infra-logic separation epic (context for the port/seam discipline this mission enforces).
  • PR #3888 — open PR touching test_layer_rules.py / test_runtime_charter_doctrine_boundary.py; hard sequencing gate for FR-013.
  • RuntimeEventEmitter disposition ADR — PR #3898, docs/adr/3.x/2026-09-06-2-runtime-event-emitter-disposition.md, currently status: Proposed (PR open); input dependency of the P4 slice (C-003, FR-011/FR-012); tasks-time gate: merged AND Accepted.
  • PR #3899 — the quick-wins PR (draft), covering the scripts/dep/symlink/prose set (incl. truststore); the src/mission_runtime residue is owned by this mission (US3/FR-014/Open Decision 8).
  • Mission A — fsm-write-path-integrity-01M1TZV6; sequencing gate C-001.
  • Missions C and D — planned-but-unminted (no kitty-specs entry, no tracker issue); the obligations recorded here are self-contained regardless.
  • Evidence chain — work/post-convergence/19–29 (reports 20, 21, 24, 25, 27 are the dossier's sources; 27 is the after-action debrief; 29 is copied into this mission at research/29-missionB-research-dossier.md). The in-mission dossier fully subsumes those session reports: the work/post-convergence/ files are session-local provenance only (gitignored, machine-local, untracked) and are NOT required to plan or implement — refinement needs only the dossier.