diff --git a/package.json b/package.json
index 34bf32f..8d841c7 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "agenthub",
-  "version": "0.7.1",
+  "version": "0.7.2",
   "description": "Local coordination layer for AI coding agents",
   "type": "module",
   "main": "./dist/index.js",
diff --git a/src/cli/index.ts b/src/cli/index.ts
index bb165c2..4845b0b 100644
--- a/src/cli/index.ts
+++ b/src/cli/index.ts
@@ -115,7 +115,7 @@ async function runRemote(serverUrl: string, fn: () => Promise<void>): Promise<vo
 export function createProgram(cwd: string): Command {
   const program = new Command('agenthub')
     .description('Local coordination layer for AI coding agents')
-    .version('0.7.1')
+    .version('0.7.2')
     .option('--server <url>', 'AgentHub server URL (env: AGENTHUB_SERVER)');
 
   program
diff --git a/src/core/templates.ts b/src/core/templates.ts
index 6f34e06..447a354 100644
--- a/src/core/templates.ts
+++ b/src/core/templates.ts
@@ -43,29 +43,24 @@ On start: \`agenthub hello --agent <you> --role architect\`, then \`agenthub wat
 function implementerMd(cliName: string, agentName: string, roles: string): string {
   return `# ${cliName} — AgentHub roles: ${roles}
 
-**On your first turn, RUN this and follow its output — do not just summarize:**
+**On your first turn, get to work — do not just summarize.**
 
-\`\`\`
-agenthub work --agent ${agentName} --role implementer
-\`\`\`
-
-\`work\` waits until a task addressed to you is ready (newly delegated OR reopened
-after review), claims it, and prints the task + its handoff. Then:
+**Preferred — the AgentHub MCP tools** (if the \`agenthub\` MCP server is registered):
+call **agenthub_work** with \`{ agent: "${agentName}", role: "implementer" }\`. It waits
+until a task addressed to you is ready (newly delegated OR reopened after review),
+claims it, and returns the task + its handoff. Then:
 
 1. Implement the task.
-2. **Submit for review (NOT done):**  \`agenthub task review <id>\`
-   and report:  \`agenthub memory add --title "<id> result" --category implementation
-   --content "<what you did / how to verify it>"\`
-3. **Run \`agenthub work --agent ${agentName} --role implementer\` again** — it blocks
-   until your next task (or a reopened one) is ready, then auto-claims it. This is
-   the loop: work → implement → review → work. Run it in the background so the wait
-   doesn't tie up your turn.
+2. **Submit for review (NOT done):** call **agenthub_task_review** \`{ id }\`, and record
+   what you did with **agenthub_memory_add** \`{ title: "<id> result", content: "…" }\`.
+3. Call **agenthub_work** again. Loop: work → implement → review → work.
 
-(\`agenthub start --agent ${agentName} --role implementer\` is the one-shot variant:
-it claims an already-open task but does not wait.)
+**CLI equivalent** (if MCP isn't set up): \`agenthub work --agent ${agentName} --role implementer\`,
+then \`agenthub task review <id>\` + \`agenthub memory add …\`, then \`agenthub work\` again
+(run it in the background so the wait doesn't tie up your turn).
 
-⚠️ NEVER run \`agenthub task done\` — only the architect approves and closes tasks.
-You drive the AgentHub CLI yourself; the human does not type these commands for you.
+⚠️ NEVER close a task yourself (no \`agenthub_task_done\` / \`agenthub task done\`) — only the
+architect approves and closes. You drive the tools yourself; the human doesn't type them for you.
 `;
 }