feat: wire up all Phase 2-6 features into daemon and config

Integrate all new features into the shared infrastructure:
- Config schema: add memory, discord, slack, process, web_search schemas
- Daemon wiring: memory store init, tool registration, channel adapters
- Orchestrator: memory injection into system prompt, extraction on compaction
- Agent: add setSystemPrompt() for dynamic prompt updates
- Channel/tool index: export new adapters and tool factories
- Add @slack/bolt, discord.js, turndown, linkedom, @mozilla/readability deps
- Update state.json with Phase 3b completion (494 tests passing)

src/backends/native/agent.ts

@@ -199,6 +199,10 @@ export class NativeAgent {
     return this.currentTier;
   }
 
+  setSystemPrompt(prompt: string): void {
+    this.systemPrompt = prompt;
+  }
+
   setOnToolUse(callback: ((event: ToolUseEvent) => void) | undefined): void {
     this.onToolUse = callback;
   }

src/backends/native/orchestrator.ts

@@ -3,6 +3,7 @@ import type { ChatRequest, Message, TokenUsage } from '../../models/types.js';
 import type { Session } from '../../session/index.js';
 import type { ToolRegistry } from '../../tools/registry.js';
 import type { ToolExecutor } from '../../tools/executor.js';
+import type { MemoryStore } from '../../memory/store.js';
 import { NativeAgent } from './agent.js';
 import type { ToolUseEvent } from './agent.js';
 import { shouldCompact } from '../../context/tokens.js';
@@ -64,6 +65,8 @@ export interface OrchestratorConfig {
   modelName?: string;
   /** Optional override for the context window size (in tokens). */
   contextWindow?: number;
+  /** Optional memory store for injecting persistent memory into the system prompt. */
+  memoryStore?: MemoryStore;
 }
 
 // ── AgentOrchestrator ─────────────────────────────────────────────────
@@ -86,6 +89,8 @@ export class AgentOrchestrator {
   private _compactionConfig?: CompactionConfig;
   private _modelName?: string;
   private _contextWindow?: number;
+  private _memoryStore?: MemoryStore;
+  private _systemPromptBase: string;
   private _usageByTier: Map<string, TierUsageStats> = new Map();
 
   constructor(config: OrchestratorConfig) {
@@ -97,6 +102,8 @@ export class AgentOrchestrator {
     this._compactionConfig = config.compaction;
     this._modelName = config.modelName;
     this._contextWindow = config.contextWindow;
+    this._memoryStore = config.memoryStore;
+    this._systemPromptBase = config.systemPrompt;
 
     // Create the primary NativeAgent for user-facing conversation
     this._agent = new NativeAgent({
@@ -178,6 +185,7 @@ export class AgentOrchestrator {
    * exceeds the context window threshold and compacts it before processing.
    */
   async process(userMessage: string): Promise<string> {
+    this._injectMemoryContext();
     await this.compactIfNeeded();
     return this._agent.process(userMessage);
   }
@@ -199,6 +207,7 @@ export class AgentOrchestrator {
       messages,
       orchestrator: this,
       config,
+      memoryStore: this._memoryStore,
     });
 
     // If nothing was actually compacted, skip the replace
@@ -268,6 +277,27 @@ export class AgentOrchestrator {
 
   // ── Private helpers ───────────────────────────────────────────────
 
+  /**
+   * Inject persistent memory context into the primary agent's system prompt.
+   * Reads from the memory store and appends relevant context to the base
+   * system prompt. If no memory store is configured or no memory content
+   * exists, restores the original base prompt.
+   */
+  private _injectMemoryContext(): void {
+    if (!this._memoryStore) {
+      return;
+    }
+
+    const memoryContext = this._memoryStore.getContextForPrompt();
+    if (!memoryContext) {
+      this._agent.setSystemPrompt(this._systemPromptBase);
+      return;
+    }
+
+    const enrichedPrompt = `${this._systemPromptBase}\n\n# Memory Context\n\nThe following is your persistent memory. Use it to maintain continuity across sessions.\n\n${memoryContext}`;
+    this._agent.setSystemPrompt(enrichedPrompt);
+  }
+
   /**
    * Check whether automatic compaction should run, and if so, compact.
    * Called before each `process()` call when compaction is configured.