feat(agent): unify models and validate OpenCode providers

Generate chat, task and learning models from agent-models.json.
Use WaterNetwork fast/expert models with 272k context and validate
upstream names before readiness. Add an isolated model check command
with signal-safe cleanup and plugin, Skill and MCP management docs.

Verified in isolation: type checks, 26 focused tests and both model
catalog checks; no model inference requests.

src/learning/orchestrator.ts

@@ -2,6 +2,7 @@ import { z } from "zod";
 
 import { writeLearningAuditLog } from "../audit/learningAudit.js";
 import { type SupportedModel } from "../chat/models.js";
+import { toRuntimeModel } from "../chat/modelConfig.js";
 import { type ChatRequestContext } from "../chat/sessionBridge.js";
 import { config } from "../config.js";
 import { type SessionTurnRecord, SessionTranscriptStore } from "../sessions/transcriptStore.js";
@@ -158,7 +159,7 @@ export class LearningOrchestrator {
       await this.runtime.prompt(
         gateSession.id,
         buildGatePrompt({ recentTurns }),
-        GATE_MODEL,
+        toRuntimeModel(config.LEARNING_GATE_MODEL),
       );
       const messages = await this.runtime.messages(gateSession.id, 20);
       const assistantMessage = [...messages]
@@ -558,25 +559,6 @@ const collectTextContent = (
     .map((part) => part.text)
     .join("");
 
-const toRuntimeModel = (model?: SupportedModel) => {
-  if (!model) {
-    return undefined;
-  }
-  const [providerID, modelID] = model.split("/");
-  if (!providerID || !modelID) {
-    return undefined;
-  }
-  return {
-    modelID,
-    providerID,
-  };
-};
-
-const GATE_MODEL = {
-  modelID: "deepseek-flash",
-  providerID: "deepseek",
-} as const;
-
 const REDACTED_AUDIT_FIELD = "[redacted by persistence policy]";
 
 const sanitizeAuditDetail = (detail?: string) => {