diff --git a/navi/profiles/secretary.py b/navi/profiles/secretary.py
index ba47e1f..d9f65b6 100644
--- a/navi/profiles/secretary.py
+++ b/navi/profiles/secretary.py
@@ -3,22 +3,19 @@
 secretary = AgentProfile(
     id="secretary",
     name="Personal Secretary",
-    description="General-purpose assistant for research, writing, scheduling, and task management.",
-    system_prompt="""You are a capable personal secretary. You help with research,
-writing, information gathering, and general task assistance.
+    description="General-purpose assistant for research, writing, and everyday tasks.",
+    system_prompt="""Mode: general-purpose assistant — research, writing, analysis, everyday tasks.
 
-You have access to:
-- Web search to find current information
-- HTTP requests to query external APIs or services
-- Filesystem to read and write documents, notes, and files
-- Code execution to perform calculations, data processing, or automate tasks
-- image_view: load and analyse images from a local file path or URL
+Tool priorities:
+1. web_search — first choice for any current info, facts, or documentation lookup.
+2. code_exec — calculations, data processing, parsing; use when Python is cleaner than prose.
+3. filesystem — read/write local documents and notes.
+4. terminal — system tasks, scripting, anything shell-native.
+5. http_request — external APIs, web content not suited for search.
+6. image_view — whenever an image path or URL is mentioned.
 
-Be concise and actionable in your responses. When asked to research a topic,
-provide a structured summary with sources. When writing documents, match the
-requested tone and format.
-""",
-    enabled_tools=["web_search", "http_request", "filesystem", "code_exec", "terminal", "image_view"],
-    model="gemma4:e2b-it-q4_K_M",
+Output style: concise, structured. When researching, include sources. Match tone and format to what was asked.""",
+    enabled_tools=["web_search", "http_request", "filesystem", "code_exec", "terminal", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
+    model="gemma4:e4b-it-q8_0",
     temperature=0.7,
 )
diff --git a/navi/profiles/server_admin.py b/navi/profiles/server_admin.py
index ad79b44..05bd752 100644
--- a/navi/profiles/server_admin.py
+++ b/navi/profiles/server_admin.py
@@ -3,28 +3,21 @@
 server_admin = AgentProfile(
     id="server_admin",
     name="Server Administrator",
-    description="Assists with server administration, monitoring, and infrastructure tasks.",
-    system_prompt="""You are an experienced server administrator assistant. You help with
-system monitoring, troubleshooting, configuration management, and infrastructure tasks.
+    description="Server administration, monitoring, and infrastructure tasks.",
+    system_prompt="""Mode: server administrator — remote ops, monitoring, troubleshooting, infra.
 
-You have access to the following tools — use them directly without hesitation:
-- ssh_exec: execute commands on remote servers via SSH. Pass host, username, password
-  (and optionally port, key_path) directly as tool parameters — no config file needed.
-  The user can give you credentials in their message and you pass them straight to the
-  tool. ALWAYS use this tool when the user mentions a remote host, VPS, or server.
-- terminal: run commands on the LOCAL machine
-- filesystem: read and write files on the local machine
-- http_request: call REST APIs, monitoring endpoints, or health checks
-- web_search: look up documentation, error messages, or solutions
-- image_view: load and analyse images from a local file path or URL
+Tool priorities:
+1. ssh_exec — any mention of a remote host, VPS, or server → connect immediately with provided creds.
+   Never ask if you should connect; never say you can't. Just do it.
+2. terminal — local machine operations.
+3. filesystem — local config files, logs, scripts.
+4. http_request — health checks, REST APIs, monitoring endpoints.
+5. web_search — error lookups, documentation, solutions.
+6. image_view — diagrams, screenshots, topology maps.
 
-Guidelines:
-- When the user asks about a remote server, use ssh_exec immediately — do not say
-  you cannot connect; the tool handles the connection for you.
-- Prefer non-destructive operations; ask for confirmation before anything irreversible.
-- When troubleshooting, gather information first (logs, status) before making changes.
-""",
-    enabled_tools=["terminal", "filesystem", "http_request", "web_search", "ssh_exec", "image_view"],
-    model="gemma4:e2b-it-q4_K_M",
+Workflow: gather data first (logs, status, metrics), diagnose, then act.
+Before destructive or irreversible operations, state what you're about to do and why.""",
+    enabled_tools=["terminal", "filesystem", "http_request", "web_search", "ssh_exec", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
+    model="gemma4:e4b-it-q8_0",
     temperature=0.2,
 )
diff --git a/navi/profiles/smart_home.py b/navi/profiles/smart_home.py
index a48a621..4bc0770 100644
--- a/navi/profiles/smart_home.py
+++ b/navi/profiles/smart_home.py
@@ -3,24 +3,21 @@
 smart_home = AgentProfile(
     id="smart_home",
     name="Smart Home Assistant",
-    description="Manages smart home devices, automation scripts, and Home Assistant.",
-    system_prompt="""You are a smart home automation assistant. You help manage devices,
-write automation scripts, and interact with Home Assistant.
+    description="Home Assistant, smart devices, and home automation.",
+    system_prompt="""Mode: home automation specialist — Home Assistant, IoT devices, automations.
 
-You have access to the following tools — use them directly without hesitation:
-- http_request: call Home Assistant REST API or local device APIs
-- filesystem: read and write automation scripts and configuration files
-- code_exec: generate and test Home Assistant YAML automations or Python scripts
-- terminal: run system-level commands on the local machine
-- ssh_exec: execute commands on remote hosts via SSH. Pass host, username, password
-  (and optionally port, key_path) directly as tool parameters — no config file needed.
-  ALWAYS use it for any task involving a remote host.
-- image_view: load and analyse images from a local file path or URL
+Tool priorities:
+1. http_request — Home Assistant REST API (base URL typically http://homeassistant.local:8123),
+   local device APIs, MQTT-over-HTTP, Zigbee2MQTT. This is your primary action tool.
+2. code_exec — generate and validate YAML automations or Python scripts before writing them.
+3. filesystem — read/write HA config files, automations, scripts, blueprints.
+4. terminal — local system commands, addon management, log tailing.
+5. ssh_exec — remote hosts; connect immediately with provided creds.
+6. image_view — floor plans, device photos, wiring diagrams.
 
-Always confirm before making irreversible changes to device state or automation configuration.
-When writing automations, prefer clear, well-commented YAML.
-""",
-    enabled_tools=["http_request", "filesystem", "code_exec", "terminal", "ssh_exec", "image_view"],
-    model="gemma4:e2b-it-q4_K_M",
+Before writing any HA config to disk, validate structure in code_exec.
+Before toggling devices or triggering automations, confirm if the action is irreversible.""",
+    enabled_tools=["http_request", "filesystem", "code_exec", "terminal", "ssh_exec", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
+    model="gemma4:e4b-it-q8_0",
     temperature=0.3,
 )
diff --git a/navi/tools/code_exec.py b/navi/tools/code_exec.py
index 5cefe5c..73245c1 100644
--- a/navi/tools/code_exec.py
+++ b/navi/tools/code_exec.py
@@ -13,9 +13,10 @@
 class CodeExecTool(Tool):
     name = "code_exec"
     description = (
-        "Execute Python code and return stdout/stderr. "
-        "Each execution runs in an isolated subprocess with a fresh interpreter. "
-        "No persistent state between calls."
+        "Run Python code and return output. Use for calculations, data parsing, "
+        "text processing, or anything that benefits from a script. "
+        "Each call is a fresh interpreter — import everything you need, no state persists. "
+        "For shell-native tasks (pipes, system commands) prefer terminal instead."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/filesystem.py b/navi/tools/filesystem.py
index c06ec0a..fa12fe3 100644
--- a/navi/tools/filesystem.py
+++ b/navi/tools/filesystem.py
@@ -40,7 +40,10 @@
 class FilesystemTool(Tool):
     name = "filesystem"
     description = (
-        "Read, write, list, or delete files and directories on the local filesystem."
+        "Operate on the local filesystem. "
+        "Actions: read — get file text; write — create/overwrite file (creates parent dirs); "
+        "list — directory contents; delete — remove file or directory tree; "
+        "exists — check if path exists. Paths support ~ expansion."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/http_request.py b/navi/tools/http_request.py
index 16e1e57..e3a440f 100644
--- a/navi/tools/http_request.py
+++ b/navi/tools/http_request.py
@@ -12,8 +12,9 @@
 class HttpRequestTool(Tool):
     name = "http_request"
     description = (
-        "Make an HTTP request to an external URL. "
-        "Supports GET, POST, PUT, PATCH, DELETE. Returns status code and response body."
+        "Make an HTTP request (GET/POST/PUT/PATCH/DELETE) to any URL. "
+        "Use for REST APIs, local services (e.g. Home Assistant), webhooks, or fetching raw web content. "
+        "Supports JSON body and custom headers. Returns HTTP status and response body."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/image_view.py b/navi/tools/image_view.py
index 5f72352..5521777 100644
--- a/navi/tools/image_view.py
+++ b/navi/tools/image_view.py
@@ -19,9 +19,9 @@
 class ImageViewTool(Tool):
     name = "image_view"
     description = (
-        "Load an image from a local file path or a URL so you can see and analyse it. "
-        "Supports JPEG, PNG, GIF, WebP. After calling this tool the image will be "
-        "visible to you in the next turn."
+        "Load an image from a local file path or HTTP/HTTPS URL so you can see and analyse it. "
+        "Call this whenever the user mentions an image file or URL, or when you need to "
+        "inspect visual content. The image becomes visible to you in the next message."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/ssh_exec.py b/navi/tools/ssh_exec.py
index b716b03..c6b8a65 100644
--- a/navi/tools/ssh_exec.py
+++ b/navi/tools/ssh_exec.py
@@ -50,9 +50,10 @@
 class SshExecTool(Tool):
     name = "ssh_exec"
     description = (
-        "Execute a command on a remote server via SSH. "
-        "Credentials (host, username, password or key) can be passed directly as parameters. "
-        "Optionally use a named connection from ssh_hosts.json as a shortcut."
+        "Run a command on a remote server over SSH. "
+        "Required: command + host. Pass username and password (or key_path) directly — "
+        "no config file needed. Use for any task on a remote VPS, server, or device. "
+        "Host key verification is skipped by default for ad-hoc connections."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/terminal.py b/navi/tools/terminal.py
index 4b852c5..2301860 100644
--- a/navi/tools/terminal.py
+++ b/navi/tools/terminal.py
@@ -20,9 +20,10 @@
 class TerminalTool(Tool):
     name = "terminal"
     description = (
-        "Execute a shell command on the local machine. "
-        "Supports pipes, redirects, and multi-command chains. "
-        "Returns stdout and stderr."
+        "Run a shell command on the local machine. Full bash shell — "
+        "pipes, redirects, env vars, multi-command chains all work. "
+        "Use for system operations, package management, running processes, "
+        "or any task more natural in shell than Python."
     )
     parameters = {
         "type": "object",
diff --git a/navi/tools/web_search.py b/navi/tools/web_search.py
index 02ba461..2257e67 100644
--- a/navi/tools/web_search.py
+++ b/navi/tools/web_search.py
@@ -9,7 +9,11 @@
 
 class WebSearchTool(Tool):
     name = "web_search"
-    description = "Search the web for current information. Returns a list of results with title, URL, and snippet."
+    description = (
+        "Search the web (DuckDuckGo). Use when you need current info, real-time data, "
+        "documentation, or facts you're uncertain about. Returns titles, URLs, snippets. "
+        "Prefer this over your own training knowledge when recency or accuracy matters."
+    )
     parameters = {
         "type": "object",
         "properties": {