Route explicit proposals into Modeling

epiphany-core/src/bin/epiphany-mvp-coordinator.rs

@@ -481,10 +481,10 @@ fn run_coordinator(args: &Args) -> Result<Value> {
             .get("coordinator")
             .cloned()
             .unwrap_or_else(|| json!({"action": "regatherManually"}));
-        let action = coordinator["action"]
-            .as_str()
-            .unwrap_or("regatherManually")
-            .to_string();
+        let action = apply_explicit_proposal_routing(
+            &mut coordinator,
+            args.proposal_modeling_request_id.as_deref(),
+        );
 
         let snapshot_name = format!("step-{index:02}-{action}.txt");
         fs::write(
@@ -1732,6 +1732,28 @@ fn now() -> String {
     chrono::Utc::now().to_rfc3339_opts(SecondsFormat::Secs, true)
 }
 
+fn apply_explicit_proposal_routing(
+    coordinator: &mut Value,
+    proposal_modeling_request_id: Option<&str>,
+) -> String {
+    let action = coordinator["action"]
+        .as_str()
+        .unwrap_or("regatherManually");
+    if action == "awaitFrontierProposal"
+        && proposal_modeling_request_id.is_some_and(|value| !value.trim().is_empty())
+    {
+        coordinator["action"] = json!("launchModeling");
+        coordinator["targetRole"] = json!("modeling");
+        coordinator["canAutoRun"] = json!(true);
+        coordinator["requiresReview"] = json!(false);
+        coordinator["reason"] = json!(
+            "A typed proposal Modeling request was supplied explicitly; launch Modeling through that request authority."
+        );
+        return "launchModeling".to_string();
+    }
+    action.to_string()
+}
+
 fn coordinator_runtime_identity_options(
     runtime_id: &str,
     created_at: String,
@@ -1849,6 +1871,34 @@ fn home_dir() -> PathBuf {
 mod tests {
     use super::*;
 
+    #[test]
+    fn explicit_typed_proposal_turns_await_into_modeling_launch() {
+        let mut coordinator = json!({
+            "action": "awaitFrontierProposal",
+            "targetRole": "imagination",
+            "canAutoRun": false,
+            "requiresReview": true,
+            "reason": "Await a proposal."
+        });
+        assert_eq!(
+            apply_explicit_proposal_routing(
+                &mut coordinator,
+                Some("repo-frontier-proposal-modeling-request")
+            ),
+            "launchModeling"
+        );
+        assert_eq!(coordinator["action"], "launchModeling");
+        assert_eq!(coordinator["targetRole"], "modeling");
+        assert_eq!(coordinator["canAutoRun"], true);
+        assert_eq!(coordinator["requiresReview"], false);
+
+        let mut idle = json!({"action": "awaitFrontierProposal"});
+        assert_eq!(
+            apply_explicit_proposal_routing(&mut idle, None),
+            "awaitFrontierProposal"
+        );
+    }
+
     #[test]
     fn coordinator_preserves_the_authenticated_runtime_identity() {
         let options = coordinator_runtime_identity_options(

notes/fresh-workspace-handoff.md

@@ -7,8 +7,8 @@ outside Codex. CultCache persists state, CultNet carries typed messages, and
 CultMesh projects inspectable surfaces. Codex remains only as the retained
 OpenAI-compatible authentication and model-transport boundary.
 
-The local `codex/epiphany-shakedown-live` branch has a focused, tested
-Mind-to-Soul launch-context repair pending commit and publication.
+The local `codex/epiphany-shakedown-live` branch has a focused, tested explicit
+proposal-routing repair pending commit and publication.
 
 Current host authority:
 
@@ -60,6 +60,12 @@ The current live readiness result is mixed:
 - Launch assembly now resolves the route's exact RepoModel admission receipt,
   its Mind review, the bound Modeling acceptance, its Mind gateway review, and
   its Mind state commit. Soul launch fails closed if that chain is incomplete.
+- Fresh v18 cold bootstrap and ordinary Modeling/Mind admission succeeded on the
+  current exact release. Its selected typed proposal then exposed a Self fault:
+  `awaitFrontierProposal` exits before consulting the explicit proposal request.
+- Fresh Windows bootstrap also requires the extended-length spelling of the
+  external workspace-coverage path; the equivalent ordinary absolute path is
+  incorrectly rejected as noncanonical.
 - Resident artifact retention is unbounded. Durable Windows supervision,
   crash/reboot recovery, session closure, long-duration resource behavior, and
   Linux/Yggdrasil cognition remain unproven.
@@ -81,11 +87,11 @@ The current live readiness result is mixed:
 
 ## Next action
 
-Commit and publish the Mind-to-Soul context repair, package the exact release,
-then run a fresh proposal-to-Modeling-to-Mind-to-Hands-to-Soul circuit and
-require Soul to judge both the admission and consequence receipt bodies. After
-that, continue the organ matrix, retention, supervised recovery, session
-closure, long-duration resource, and Linux/Yggdrasil proofs.
+Commit and publish explicit proposal routing, then resume v18 using its existing
+selected proposal and require Modeling, Mind, Hands, and Soul to consume one
+exact typed chain. After that, continue the organ matrix, Windows cold-start
+path repair, retention, supervised recovery, session closure, long-duration
+resource, and Linux/Yggdrasil proofs.
 
 ## Immediate Re-entry Instruction
 

state/map.yaml

@@ -34,9 +34,9 @@ invariants:
   - Evidence and git retain history; this map retains only state that changes future decisions.
 
 current_status:
-  summary: Epiphany is awake enough for supervised Starfire shakedown, not deployment-ready. The live typed path reached proposal-bound Modeling, Mind admission, a scoped Hands gate, native Hands receipts, and an admitted Soul needs-evidence verdict. Soul verified the Hands consequence but correctly refused the end-to-end claim because Self's launch packet omitted the Mind admission bodies. Launch assembly is being repaired to carry the exact route-bound RepoModel review and receipt, Modeling acceptance, Mind gateway review, and Mind state commit beside the Hands chain. Resident artifact growth remains unbounded, durable Windows supervision and crash/reboot recovery are unproven, and no Linux or Yggdrasil cognition run has been performed.
+  summary: Epiphany is awake enough for supervised Starfire shakedown, not deployment-ready. A fresh v18 Body proved authenticated cold bootstrap, canonical Mind seeding, repository observation, ordinary Modeling, and Mind admission on the current exact release. The next typed proposal exposed a Self routing fault: `awaitFrontierProposal` stopped before consulting an explicitly supplied proposal Modeling request. That boundary is under repair. The earlier live path reached scoped Hands receipts and an admitted Soul needs-evidence verdict; the repaired admission-plus-consequence Soul packet still needs a fresh circuit. Resident artifact growth, ordinary Windows path ergonomics, durable supervision/recovery, and Linux/Yggdrasil cognition remain open.
   operational_state:
-    local_worktree: mind_to_soul_context_repair_under_verification
+    local_worktree: explicit_proposal_routing_repair_under_verification
     operator_wireguard: restored
     yggdrasil: canonical_small_public_crossing_body
     starfire: temporary_epiphany_cognition_body
@@ -52,8 +52,8 @@ current_status:
     bifrost_feedback_crossing: authenticated_transport_only
     yggdrasil_epiphany_services: absent
     resident_physiology: heartbeat_and_self_running
-  unresolved_authority: The full route remains unproven until a fresh Soul launch receives and judges both the route-bound Mind admission chain and the native Hands consequence chain; deployment authority remains absent.
-  next_action: Commit and publish the Mind-to-Soul context repair, then run a fresh exact-release proposal-to-Soul circuit and require a verdict grounded in both admission and consequence receipts.
+  unresolved_authority: Self currently stops at awaitFrontierProposal before using an explicit typed proposal request; the full route and deployment authority remain unproven.
+  next_action: Commit and publish explicit proposal routing, then resume v18 with its already-selected proposal and require Modeling, Mind, Hands, and Soul to consume one exact typed chain.
 
 active_subgoals:
   - Prove Persona, Self, Eyes, Modeling, Imagination, Hands, Soul, Mind, Continuity, and nervous-system behavior individually.
@@ -63,7 +63,7 @@ active_subgoals:
 
 organ_readiness:
   persona: partial; authenticated feedback transport is live, but public cognition and speech consequence still need a fresh end-to-end proof.
-  self: partial; resident routing and the live frontier route work, but the repaired admission-to-Soul packet needs a fresh exact-release run.
+  self: partial; resident routing and the live frontier route work, while explicit proposal selection currently needs a repaired transition from waiting to Modeling launch.
   eyes: unproven in the current live circuit; coordinator status still reports no accepted Research result.
   modeling: partial; real source-grounded proposal-bound Modeling and Mind admission are proven, including canonical claim grounding.
   imagination: unproven in the current campaign; the planning substrate is empty.
@@ -71,7 +71,7 @@ organ_readiness:
   soul: partial; a live needs-evidence refusal is proven, but no complete admission-plus-consequence verdict exists yet.
   mind: partial; RepoModel and thread-state admission work, while complete receipt projection into Soul is awaiting a fresh live proof.
   continuity: partial; heartbeat and resident Self survive sequential cycles, while restart, crash, reboot, closure, and retention remain open.
-  nervous_system: partial; detached scheduling, runtime jobs, typed feedback, and operator projection work, while long-duration pressure and resource plateau remain open.
+  nervous_system: partial; detached scheduling, runtime jobs, typed feedback, and operator projection work, while ordinary Windows coverage-store paths, long-duration pressure, and resource plateau remain open.
 
 forbidden_actions:
   - Do not issue a speculative deployment grant.