diff --git a/manuals/spawn_agent.md b/manuals/spawn_agent.md
index 57f7b77..a16cb4d 100644
--- a/manuals/spawn_agent.md
+++ b/manuals/spawn_agent.md
@@ -5,21 +5,33 @@
 
 **One plan step = one spawn_agent call.** If your plan has three AGENT steps, make three separate calls.
 
-**CRITICAL: spawn_agent is SYNCHRONOUS.** It blocks until the sub-agent fully completes (or times out after 5 minutes).
+**SYNCHRONOUS** — blocks until the sub-agent fully completes or times out (5 minutes hard limit).
 
 ## Parameters
 
 | Parameter | Required | Description |
 |-----------|----------|-------------|
 | `task` | yes | Goal for this one step, success criteria, expected output format. End with: "Complete ALL assigned work before responding. Your output is final." |
-| `briefing` | no | Credentials, IPs, file paths, constraints, step-by-step instructions — injected as system-level context into the sub-agent. |
+| `briefing` | no | Credentials, IPs, file paths, constraints, step-by-step instructions — injected into the sub-agent's system prompt as `## Task context`. |
 | `profile_id` | no | Which profile to use (`secretary`, `server_admin`, `developer`). Defaults to current session's profile. |
-| `system_prompt` | no | Role specialisation for this task (e.g. "You are a security auditor. Report findings by severity."). Injected on top of the profile's built-in subagent prompt. |
-| `max_iterations` | no | Tool-call iteration limit (default: 20). |
+| `system_prompt` | no | Role specialisation injected into the sub-agent's system prompt between the executor persona and the briefing (e.g. "You are a security auditor. Report findings by severity."). |
+| `max_iterations` | no | Tool-call iteration limit (default: **40**). |
+
+## Sub-agent system prompt structure
+
+The sub-agent receives a completely separate system prompt — no persona, no orchestrator instructions, no available-profiles block. It is built from up to three parts (separated by `---`):
+
+```
+1. profile.subagent_system_prompt   ← focused executor persona (from subagent_system_prompt.txt)
+2. system_prompt param              ← role specialisation (if provided)
+3. ## Task context\n\n{briefing}    ← credentials, instructions (if briefing provided)
+```
+
+Fallback: if the profile has no `subagent_system_prompt.txt`, uses `profile.system_prompt`.
 
 ## Sub-agent tools
 
-Sub-agents receive a focused tool set (no todo, switch_profile, email_manager, etc.):
+Sub-agents receive a dedicated, focused tool set (defined in `subagent_tools` in profile config):
 
 | Profile | Sub-agent tools |
 |---------|----------------|
@@ -27,44 +39,44 @@
 | `server_admin` | scratchpad, reflect, web_search, http_request, filesystem, code_exec, terminal, ssh_exec, image_view, share_file |
 | `developer` | scratchpad, reflect, web_search, web_view, http_request, filesystem, code_exec, terminal, image_view, reload_tools, test_tool, share_file |
 
+`spawn_agent` is always excluded — recursion is impossible.
+
 ## Result format
 
-The result header tells you how the sub-agent finished:
+The result always starts with a header visible only to you (never repeat it to the user):
 - `[Sub-agent completed ...]` — finished normally; synthesise the findings.
-- `[Sub-agent hit iteration limit ...]` — may be incomplete; note what's missing in your response.
+- `[Sub-agent hit iteration limit ...]` — may be incomplete; note what's missing.
 
-**Never repeat the result header to the user.** Synthesise findings in your own words.
-
-## Example: correct multi-agent execution
+## Multi-agent execution pattern
 
 ```
-Plan step 2 → AGENT  →  spawn_agent(task="Research pricing for product X ...", briefing="...")
-Plan step 3 → AGENT  →  spawn_agent(task="Research pricing for product Y ...", briefing="...")
+Plan step 2 → AGENT  →  spawn_agent(task="Research pricing for X ...", briefing="...")
+Plan step 3 → AGENT  →  spawn_agent(task="Research pricing for Y ...", briefing="...")
 Plan step 4 → SELF   →  synthesise both results, write final answer
 ```
 
-**Wrong:** `spawn_agent(task="Research X and Y pricing and write a comparison")` — two steps, one call.
+**Wrong:** `spawn_agent(task="Research X and Y and compare")` — two steps, one call.
 
 ## Full example
 
 ```json
 {
   "task": "Check CPU temperature and memory usage. Return a table: metric, value, unit, status (ok/warn/crit). Complete ALL assigned work before responding. Your output is final.",
-  "briefing": "Host: 192.168.1.75\nUser: gmikcon\nPassword: getroot\nUse ssh_exec. Check temperature via 'sensors' or /sys/class/thermal. Check memory via 'free -h'.",
+  "briefing": "Host: 192.168.1.75\nUser: gmikcon\nPassword: getroot\nUse ssh_exec. Check temperature via 'sensors' or /sys/class/thermal/thermal_zone*/temp (divide by 1000). Check memory via 'free -h'.",
   "profile_id": "server_admin",
-  "system_prompt": "You are a system metrics collector. Report all values in a structured table."
+  "system_prompt": "You are a system metrics collector. Report all values in a structured table with columns: metric, value, unit, status."
 }
 ```
 
 ## After the result arrives
 
-The user cannot see sub-agent output — you must present the findings yourself.
+The user cannot see sub-agent output — present findings yourself.
 
-1. Read the result header to know if it completed or hit the limit.
-2. Extract key findings and present them clearly to the user.
-3. If incomplete, decide: retry with a more focused task, or handle inline.
+1. If "hit iteration limit" — note what is missing in your response.
+2. Synthesise key findings in your own words.
+3. If result is insufficient, spawn again with a more focused task.
 
 ## What the sub-agent cannot do
-- Spawn further sub-agents (recursion is blocked)
-- Access your conversation history
-- Use administrative tools: todo, switch_profile, list_profiles, email_manager, delete_tool
+- Spawn further sub-agents (recursion blocked)
+- Access conversation history (only context from the current call)
+- Use: todo, switch_profile, list_profiles, email_manager, delete_tool, list_tools, tool_manual
diff --git a/navi/api/websocket.py b/navi/api/websocket.py
index aafc556..872bbfc 100644
--- a/navi/api/websocket.py
+++ b/navi/api/websocket.py
@@ -110,9 +110,9 @@
     if isinstance(event, StreamStopped):
         return {"type": "stream_stopped"}
     if isinstance(event, PlanningStatus):
-        return {"type": "planning_status", "phase": event.phase, "label": event.label}
+        return {"type": "planning_status", "phase": event.phase, "label": event.label, "is_subagent": event.is_subagent}
     if isinstance(event, PlanReady):
-        return {"type": "plan_ready", "plan": event.plan}
+        return {"type": "plan_ready", "plan": event.plan, "is_subagent": event.is_subagent}
     return None
 
 
diff --git a/navi/core/agent.py b/navi/core/agent.py
index af5915e..79512ce 100644
--- a/navi/core/agent.py
+++ b/navi/core/agent.py
@@ -335,6 +335,7 @@
             async for _ev in self._run_planning(
                 context, profile, llm, mem, tool_schemas,
                 system_prompt_override=subagent_sys_msg.content,
+                is_subagent=True,
             ):
                 if isinstance(_ev, AIHelperTokensUsed):
                     pass  # token accounting only, not forwarded
@@ -750,6 +751,7 @@
         tool_schemas: list | None = None,
         messages: "list[Message] | None" = None,
         system_prompt_override: str | None = None,
+        is_subagent: bool = False,
     ):
         """
         Three-phase planning (async generator):
@@ -792,7 +794,7 @@
         _stop = current_stop_event.get()
 
         # ── Phase 1: Task analysis (with reasoning) ────────────────────────────
-        yield PlanningStatus(phase=1, label="Analysing task...")
+        yield PlanningStatus(phase=1, label="Analysing task...", is_subagent=is_subagent)
         _base_sys = system_prompt_override if system_prompt_override is not None else self._build_system_prompt(profile)
         phase1_system = Message(
             role="system",
@@ -847,7 +849,7 @@
             return
 
         # ── Phase 2: Execution plan ────────────────────────────────────────────
-        yield PlanningStatus(phase=2, label="Building execution plan...")
+        yield PlanningStatus(phase=2, label="Building execution plan...", is_subagent=is_subagent)
         phase2_system = Message(
             role="system",
             content=(
@@ -923,7 +925,7 @@
             return
 
         # ── Phase 3: AIHelper plan critic ──────────────────────────────────────
-        yield PlanningStatus(phase=3, label="Reviewing plan...")
+        yield PlanningStatus(phase=3, label="Reviewing plan...", is_subagent=is_subagent)
         # Independent pass: validates and corrects executor assignments against
         # the actual tool list. Falls back to Phase 2 plan on any failure.
         if tool_names_list:
@@ -990,7 +992,7 @@
         if messages is not None:
             messages.append(Message(role="assistant", content=plan_text, is_plan=True))
         log.debug("agent.plan_ready", phases=3, length=len(plan_text))
-        yield PlanReady(plan=plan_text)
+        yield PlanReady(plan=plan_text, is_subagent=is_subagent)
 
     async def _run_workers(
         self,
diff --git a/navi/core/events.py b/navi/core/events.py
index 48c38f4..7636058 100644
--- a/navi/core/events.py
+++ b/navi/core/events.py
@@ -82,10 +82,12 @@
 
     phase: 1 = Analysis, 2 = Execution plan, 3 = Plan review (AIHelper critic).
     label: short human-readable description shown next to the spinner.
+    is_subagent: True when emitted from inside run_ephemeral (subagent planning).
     """
 
     phase: int
     label: str
+    is_subagent: bool = False
 
 
 @dataclass
@@ -94,9 +96,11 @@
 
     The plan text has already been injected into session.context as an assistant
     message so the LLM will see it and follow it during execution.
+    is_subagent: True when emitted from inside run_ephemeral (subagent planning).
     """
 
     plan: str
+    is_subagent: bool = False
 
 
 @dataclass
diff --git a/navi/tools/spawn_agent.py b/navi/tools/spawn_agent.py
index 0c46d00..614d0d0 100644
--- a/navi/tools/spawn_agent.py
+++ b/navi/tools/spawn_agent.py
@@ -63,7 +63,7 @@
             },
             "max_iterations": {
                 "type": "integer",
-                "description": "Maximum tool-call iterations for the sub-agent (default: 20).",
+                "description": "Maximum tool-call iterations for the sub-agent (default: 40).",
             },
         },
         "required": ["task"],
@@ -91,7 +91,7 @@
         task = params["task"].strip()
         briefing = (params.get("briefing") or "").strip() or None
         custom_system_prompt = (params.get("system_prompt") or "").strip() or None
-        max_iterations = int(params.get("max_iterations") or 20)
+        max_iterations = int(params.get("max_iterations") or 40)
 
         # task → user message (what to do)
         # briefing → system prompt (context, credentials, instructions)
diff --git a/webclient/src/components/messages/ToolCard.vue b/webclient/src/components/messages/ToolCard.vue
index ccb1707..94dbf21 100644
--- a/webclient/src/components/messages/ToolCard.vue
+++ b/webclient/src/components/messages/ToolCard.vue
@@ -27,10 +27,23 @@
         <div class="tool-section-label">Result</div>
         <pre class="tool-code">{{ formatResult(tool.result) }}</pre>
       </div>
+      <!-- Subagent planning indicator (while subagent is in planning phase) -->
+      <div v-if="tool.planningLabel != null" class="subagent-planning-indicator">
+        <span class="spinner"></span>
+        <span class="planning-label">{{ tool.planningLabel }}</span>
+      </div>
       <!-- Subagent steps nested inside spawn_agent card -->
       <div v-if="tool.steps?.length" class="subagent-steps">
         <template v-for="(step, i) in tool.steps" :key="i">
           <ThinkingCard v-if="step.kind === 'turn_thinking'" :msg="step" />
+          <details v-else-if="step.kind === 'plan'" class="plan-card" open>
+            <summary>
+              <i class="ph ph-map-trifold"></i>
+              Plan
+              <i class="ph ph-caret-down plan-chevron"></i>
+            </summary>
+            <div class="plan-body" v-html="renderMarkdown(step.text)" />
+          </details>
           <SubagentStep v-else :tool="step" />
         </template>
       </div>
@@ -42,6 +55,7 @@
 import { ref, computed, watch } from 'vue'
 import SubagentStep from './SubagentStep.vue'
 import ThinkingCard from './ThinkingCard.vue'
+import { renderMarkdown } from '@/composables/useMarkdown.js'
 
 const props = defineProps({ tool: { type: Object, required: true } })
 
diff --git a/webclient/src/stores/chat.js b/webclient/src/stores/chat.js
index 98a9af0..4077b55 100644
--- a/webclient/src/stores/chat.js
+++ b/webclient/src/stores/chat.js
@@ -138,12 +138,24 @@
   function onPlanningStatus(data) {
     const msg = streamingMsg.value
     if (!msg) return
+    if (data.is_subagent) {
+      const spawn = _lastSpawnCard(msg)
+      if (spawn) { spawn.planningLabel = data.label ?? ''; return }
+    }
     msg.statusLabel = data.label ?? ''
   }
 
   function onPlanReady(data) {
     const msg = streamingMsg.value
     if (!msg) return
+    if (data.is_subagent) {
+      const spawn = _lastSpawnCard(msg)
+      if (spawn) {
+        spawn.planningLabel = null
+        spawn.steps.push({ kind: 'plan', text: data.plan ?? '' })
+        return
+      }
+    }
     msg.statusLabel = null  // planning done, clear the status label
     msg.tools.push({ kind: 'plan', text: data.plan ?? '' })
   }
diff --git a/webclient/src/styles/app.scss b/webclient/src/styles/app.scss
index 0ba7cd2..142b097 100644
--- a/webclient/src/styles/app.scss
+++ b/webclient/src/styles/app.scss
@@ -855,6 +855,17 @@
   background: $surface-panel-muted;
 }
 
+.subagent-planning-indicator {
+  display: flex;
+  align-items: center;
+  gap: $space-sm;
+  padding: $space-xs $space-md;
+  font-size: 12px;
+  font-style: italic;
+  color: $color-text-dark;
+  background: $surface-panel-muted;
+}
+
 // ─── Summary / compression notice ──────────────────────────────────────────
 .summary-card {
   border: $border-width-base $border-style-base $color-info;