Security Audit
aAAaqwq/AGI-Super-Team
Comprehensive security auditing for Clawdbot deployments. An agent skill from aAAaqwq/AGI-Super-Team.
Security hardening and best practices for robotic systems, covering SROS2 DDS security, network segmentation, secrets management, secure boot, and the physical-cyber safety intersection.
The automated check flagged lines worth reading first. See the safety section below.
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install arpitg1304/robotics-agent-skills robotics-security --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/robotics-security .claude/skills/robotics-security && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .claude/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-securityType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install arpitg1304/robotics-agent-skills robotics-security --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/robotics-security .agents/skills/robotics-security && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .agents/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install arpitg1304/robotics-agent-skills robotics-security --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/robotics-security .cursor/skills/robotics-security && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .cursor/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/arpitg1304/robotics-agent-skills.git --path skills/robotics-security--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install arpitg1304/robotics-agent-skills robotics-security --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/robotics-security .gemini/skills/robotics-security && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .gemini/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install arpitg1304/robotics-agent-skills robotics-securityInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/robotics-security .github/skills/robotics-security && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .github/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install arpitg1304/robotics-agent-skills robotics-security --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/arpitg1304/robotics-agent-skills.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/robotics-security .opencode/skills/robotics-security && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "robotics-security" agent skill from https://github.com/arpitg1304/robotics-agent-skills/tree/main/skills/robotics-security into .opencode/skills/robotics-security/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "robotics-security", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
robotics-securitySecurity hardening and best practices for robotic systems, covering SROS2 DDS security, network segmentation, secrets management, secure boot, and the physical-cyber safety intersection.
Robotics Security is an agent skill from arpitg1304/robotics-agent-skills. Security hardening and best practices for robotic systems, covering SROS2 DDS security, network segmentation, secrets management, secure boot, and the physical-cyber safety intersection. Use this skill when securing ROS2 communications, configuring DDS encryption and access control, hardening robot onboard computers, managing certificates and credentials, setting up network segmentation for robot fleets, or addressing the unique security challenges where cyber vulnerabilities become physical safety risks. Trigger…
Its SKILL.md is about 7.8k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.
It sits in DevOps & Cloud, covering Authorization and RBAC, Network security and Secrets management. The repository describes itself as: Agent skills that make AI coding assistants write production-grade robotics software. ROS1, ROS2, design patterns, SOLID principles, and testing — for Claude Code, Cursor… The licence is Apache-2.0.
8 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit f9bc546. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Shell commands in SKILL.md call:
apttrivydockergitgitleaksFrom the folder's file list and the shell code blocks in SKILL.md.
Hosts in commands or code, which the agent is likely to contact:
w3.orgFrom URLs in SKILL.md, links to its own repository left out.
Names these keys or tokens, usually read from environment variables:
FLEET_API_KEYAWS_SECRET_KEYCOSIGN_PRIVATE_KEYFrom names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Robotics Security loads about 7.8k tokens when it runs. Until then it costs about 209 tokens; SKILL.md has 978 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found patterns that need a careful read before installing.
sudo systemctl restart sshdssh-keygen -t ed25519 -f ~/.ssh/robot_$(hostname) -C "admin@$(hostname)"ssh-copy-id -i ~/.ssh/robot_$(hostname).pub -p 2222 robot-admin@10.10.30.5sudo apt install fail2ban -y && sudo systemctl enable --now fail2bansudo apt install unattended-upgrades -ysudo dpkg-reconfigure -plow unattended-upgradessudo apt purge -y avahi-daemon cups snapd modemmanager bluetooth bluezsudo apt autoremove -ysudo chown root:robot /etc/robot/secrets.envsudo chmod 640 /etc/robot/secrets.envAutomated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from arpitg1304/robotics-agent-skills at commit f9bc546, republished under its Apache-2.0 licence (© arpitg1304). 978 words, ~7,788 tokens.
.claude/skills/robotics-security/SKILL.md (or your agent's skills folder).Robots are unique: cyber vulnerabilities become physical threats.
NETWORK MIDDLEWARE APPLICATION
┌────────────────┐ ┌────────────────┐ ┌────────────────┐
│ Open DDS ports │───────▶│ Unauthenticated│──────────▶│ Hardcoded │
│ (7400-7500) │ │ /cmd_vel pub │ │ credentials │
│ Unsegmented LAN│ │ No msg signing │ │ Unvalidated cmd│
└────────────────┘ └────────────────┘ └────────────────┘
PHYSICAL FIRMWARE SUPPLY CHAIN
┌────────────────┐ ┌────────────────┐ ┌────────────────┐
│ USB/debug ports│───────▶│ Unsigned │──────────▶│ Compromised │
│ Serial consoles│ │ firmware OTA │ │ ROS packages │
│ Exposed SBCs │ │ No secure boot │ │ Unverified imgs│
└────────────────┘ └────────────────┘ └────────────────┘| Vector | Impact |
|---|---|
Unauthenticated /cmd_vel | Robot moves unexpectedly — injury/damage |
Sensor spoofing (/scan, /camera/image) | Robot collides, wrong decisions |
| Open DDS multicast discovery | Full topic graph enumeration by passive listener |
| USB/serial physical access | Root shell, firmware flash, data exfiltration |
| Unsigned firmware update | Persistent backdoor in motor controllers |
SROS2 wraps DDS Security to provide authentication, encryption, and access control at the DDS layer.
export ROS_SECURITY_KEYSTORE=~/sros2_keystore
ros2 security create_keystore ${ROS_SECURITY_KEYSTORE}
# Generate per-node enclaves (use exact fully-qualified node names)
ros2 security create_enclave ${ROS_SECURITY_KEYSTORE} /my_robot/camera_driver
ros2 security create_enclave ${ROS_SECURITY_KEYSTORE} /my_robot/navigation
ros2 security create_enclave ${ROS_SECURITY_KEYSTORE} /my_robot/motor_controller
ros2 security create_enclave ${ROS_SECURITY_KEYSTORE} /my_robot/teleop
# Result:
# sros2_keystore/
# ├── enclaves/my_robot/{camera_driver,navigation,...}/
# │ ├── cert.pem, key.pem # Node identity
# │ ├── governance.p7s # Signed governance
# │ └── permissions.p7s # Signed permissions
# ├── public/ca.cert.pem # CA certificate
# └── private/ca.key.pem # CA private key — PROTECT THISGovernance — domain-wide security behavior:
<?xml version="1.0" encoding="UTF-8"?>
<dds xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:noNamespaceSchemaLocation="omg_shared_ca_governance.xsd">
<domain_access_rules>
<domain_rule>
<domains><id_range><min>0</min><max>230</max></id_range></domains>
<allow_unauthenticated_participants>false</allow_unauthenticated_participants>
<enable_join_access_control>true</enable_join_access_control>
<discovery_protection_kind>ENCRYPT</discovery_protection_kind>
<liveliness_protection_kind>ENCRYPT</liveliness_protection_kind>
<rtps_protection_kind>ENCRYPT</rtps_protection_kind>
<topic_access_rules>
<topic_rule>
<topic_expression>*</topic_expression>
<enable_discovery_protection>true</enable_discovery_protection>
<enable_read_access_control>true</enable_read_access_control>
<enable_write_access_control>true</enable_write_access_control>
<metadata_protection_kind>ENCRYPT</metadata_protection_kind>
<data_protection_kind>ENCRYPT</data_protection_kind>
</topic_rule>
</topic_access_rules>
</domain_rule>
</domain_access_rules>
</dds>Permissions — per-enclave publish/subscribe rules:
<?xml version="1.0" encoding="UTF-8"?>
<dds xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:noNamespaceSchemaLocation="omg_shared_ca_permissions.xsd">
<permissions>
<grant name="/my_robot/motor_controller">
<subject_name>CN=/my_robot/motor_controller</subject_name>
<validity><not_before>2024-01-01T00:00:00</not_before>
<not_after>2035-01-01T00:00:00</not_after></validity>
<allow_rule>
<domains><id>0</id></domains>
<publish><topics><topic>rt/joint_states</topic></topics></publish>
<subscribe><topics><topic>rt/cmd_vel</topic></topics></subscribe>
</allow_rule>
<default>DENY</default>
</grant>
<grant name="/my_robot/teleop">
<subject_name>CN=/my_robot/teleop</subject_name>
<validity><not_before>2024-01-01T00:00:00</not_before>
<not_after>2035-01-01T00:00:00</not_after></validity>
<allow_rule>
<domains><id>0</id></domains>
<publish><topics><topic>rt/cmd_vel</topic></topics></publish>
<subscribe><topics><topic>rt/joy</topic></topics></subscribe>
</allow_rule>
<default>DENY</default>
</grant>
</permissions>
</dds>Expiry is a silent outage. Once not_after passes, every participant using that
grant is rejected at discovery time. Under ROS_SECURITY_STRATEGY=Enforce the robot
comes up with nodes that never see each other, and the only clue is a discovery-time
rejection in the DDS log -- no topic error, no crash. Track permissions expiry the way
you track TLS certs: alert well before the date, and rehearse the reissue procedure.
import os
from launch import LaunchDescription
from launch_ros.actions import Node
def generate_launch_description():
security_env = {
'ROS_SECURITY_KEYSTORE': os.path.expanduser('~/sros2_keystore'),
'ROS_SECURITY_ENABLE': 'true',
'ROS_SECURITY_STRATEGY': 'Enforce', # Enforce=reject unauth, Permissive=warn only
}
return LaunchDescription([
Node(package='my_robot_drivers', executable='motor_controller',
name='motor_controller', namespace='my_robot',
additional_env=security_env),
Node(package='my_robot_nav', executable='navigation',
name='navigation', namespace='my_robot',
additional_env=security_env),
])Always use Enforce in production. Permissive logs violations but allows them — debugging aid only.
Design with least privilege:
| Node | Publishes | Subscribes | Rationale |
|---|---|---|---|
motor_controller | /joint_states | /cmd_vel | Driver acts on velocity only |
navigation | /cmd_vel, /path | /scan, /odom, /map | Nav reads sensors, writes commands |
camera_driver | /camera/image_raw | (none) | Pure source — no subscriptions |
teleop | /cmd_vel | /joy | Joystick passthrough — minimal surface |
A compromised camera_driver cannot publish to /cmd_vel — permissions deny it at the DDS layer.
┌───────────────────┬──────────────────┬────────────────────────┐
│ CONTROL PLANE │ DATA PLANE │ MANAGEMENT PLANE │
│ VLAN 10 │ VLAN 20 │ VLAN 30 │
│ 10.10.10.0/24 │ 10.10.20.0/24 │ 10.10.30.0/24 │
├───────────────────┼──────────────────┼────────────────────────┤
│ /cmd_vel, /odom │ /camera/image │ SSH, Prometheus │
│ /joint_states │ /pointcloud │ Log collection │
│ /e_stop │ /map, /rosbag │ Fleet mgmt API │
├───────────────────┼──────────────────┼────────────────────────┤
│ LOW LATENCY │ HIGH BANDWIDTH │ RESTRICTED ACCESS │
│ QoS: RELIABLE │ QoS: BEST_EFFORT │ Jump host / VPN + 2FA │
└───────────────────┴──────────────────┴────────────────────────┘Management plane is never reachable from data plane. Control plane traffic never transits WiFi.
#!/bin/bash
# firewall_ros2.sh — adapt interface names to your hardware
iptables -F && iptables -X
# Default: drop inbound, allow outbound
iptables -P INPUT DROP
iptables -P FORWARD DROP
iptables -P OUTPUT ACCEPT
iptables -A INPUT -i lo -j ACCEPT # Loopback (intra-process DDS)
iptables -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT # Existing connections
iptables -A INPUT -p udp --dport 7400:7500 -s 10.10.10.0/24 -j ACCEPT # DDS discovery — control VLAN
iptables -A INPUT -p udp --dport 7500:7700 -s 10.10.10.0/24 -j ACCEPT # DDS user traffic
iptables -A INPUT -p tcp --dport 22 -s 10.10.30.0/24 -j ACCEPT # SSH — mgmt VLAN only
iptables -A INPUT -i wlan0 -d 239.255.0.0/16 -j DROP # Block multicast on WiFi
iptables -A INPUT -j LOG --log-prefix "DROPPED: " --log-level 4
iptables -A INPUT -j DROP
iptables-save > /etc/iptables/rules.v4# /etc/netplan/01-robot-vlans.yaml
network:
version: 2
renderer: networkd
ethernets:
eth0: {dhcp4: false}
vlans:
vlan10:
id: 10
link: eth0
addresses: [10.10.10.5/24]
vlan20:
id: 20
link: eth0
addresses: [10.10.20.5/24]
vlan30:
id: 30
link: eth0
addresses: [10.10.30.5/24]
routes: [{to: default, via: 10.10.30.1}]Multicast auto-discovery exposes the full topic graph. Use unicast peer lists.
<!-- cyclonedds_secure.xml -->
<CycloneDDS>
<Domain>
<General><AllowMulticast>false</AllowMulticast></General>
<Discovery>
<Peers>
<Peer address="10.10.10.1"/>
<Peer address="10.10.10.2"/>
<Peer address="10.10.10.3"/>
</Peers>
<ParticipantIndex>auto</ParticipantIndex>
</Discovery>
</Domain>
</CycloneDDS>export CYCLONEDDS_URI=file:///etc/ros2/cyclonedds_secure.xml
export RMW_IMPLEMENTATION=rmw_cyclonedds_cppFastDDS equivalent — set initialPeersList with explicit unicast locators and omit multicast locators in the participant profile. Use FASTRTPS_DEFAULT_PROFILES_FILE env var to load.
# /etc/ssh/sshd_config
Port 2222
PermitRootLogin no
PasswordAuthentication no
PubkeyAuthentication yes
AuthorizedKeysFile .ssh/authorized_keys
MaxAuthTries 3
ClientAliveInterval 300
ClientAliveCountMax 2
AllowUsers robot-admin
X11Forwarding no
AllowTcpForwarding no
PermitTunnel nosudo systemctl restart sshd
# Per-robot key pair (on management workstation)
ssh-keygen -t ed25519 -f ~/.ssh/robot_$(hostname) -C "admin@$(hostname)"
ssh-copy-id -i ~/.ssh/robot_$(hostname).pub -p 2222 robot-admin@10.10.30.5# /etc/fail2ban/jail.local
[sshd]
enabled = true
port = 2222
filter = sshd
logpath = /var/log/auth.log
maxretry = 3
bantime = 3600
findtime = 600sudo apt install fail2ban -y && sudo systemctl enable --now fail2bansudo apt install unattended-upgrades -y
sudo dpkg-reconfigure -plow unattended-upgrades
# Key settings in /etc/apt/apt.conf.d/50unattended-upgrades:
# Allowed-Origins: "${distro_id}:${distro_codename}-security"
# Automatic-Reboot: "false" # NEVER auto-reboot a running robot# Remove unnecessary packages from robot computers
sudo apt purge -y avahi-daemon cups snapd modemmanager bluetooth bluez
sudo apt autoremove -y# BAD:
class FleetClient:
def __init__(self):
self.api_key = "sk-live-abc123xyz789"# GOOD:
import os
class FleetClient:
def __init__(self):
self.api_key = os.environ['FLEET_API_KEY']# BAD: credentials in params.yaml tracked by git
fleet_manager:
ros__parameters:
aws_secret_key: "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"# GOOD: reference environment variables
fleet_manager:
ros__parameters:
aws_secret_key: "$(env AWS_SECRET_KEY)"# /etc/systemd/system/robot-nav.service
[Service]
User=robot
Group=robot
EnvironmentFile=/etc/robot/secrets.env
ExecStart=/opt/ros/humble/bin/ros2 launch my_robot nav.launch.py
Restart=always# /etc/robot/secrets.env
FLEET_API_KEY=sk-live-actual-key-here
ROS_SECURITY_KEYSTORE=/opt/robot/sros2_keystore
# Lock it down
sudo chown root:robot /etc/robot/secrets.env
sudo chmod 640 /etc/robot/secrets.env#!/bin/bash
# rotate_certs.sh — run via cron monthly
set -euo pipefail
KEYSTORE="/opt/robot/sros2_keystore"
cp -r "${KEYSTORE}" "${KEYSTORE}_backup_$(date +%Y%m%d)"
for enclave in motor_controller navigation camera_driver teleop; do
ros2 security create_enclave "${KEYSTORE}" "/my_robot/${enclave}"
done
sudo systemctl restart robot-*.service
echo "Certificates rotated at $(date)"# /etc/cron.d/robot-cert-rotation
0 3 1 * * root /opt/robot/scripts/rotate_certs.sh >> /var/log/cert-rotation.log 2>&1sudo chown -R root:robot /opt/robot/sros2_keystore
sudo find /opt/robot/sros2_keystore -type d -exec chmod 750 {} \;
sudo find /opt/robot/sros2_keystore -type f -exec chmod 640 {} \;
# CA private key — root only
sudo chmod 600 /opt/robot/sros2_keystore/private/ca.key.pem
sudo chown root:root /opt/robot/sros2_keystore/private/ca.key.pemFROM ros:humble-ros-base AS runtime
RUN apt-get update && apt-get install -y --no-install-recommends \
ros-humble-nav2-bringup && rm -rf /var/lib/apt/lists/*
RUN groupadd -g 1000 robot && useradd -u 1000 -g robot -m -s /bin/false robot
COPY --from=builder /opt/ros2_ws/install /opt/ros2_ws/install
USER robot:robot
ENTRYPOINT ["/ros_entrypoint.sh"]
CMD ["ros2", "launch", "my_robot", "nav.launch.py"]FROM ros:humble-desktop AS builder
WORKDIR /opt/ros2_ws
COPY src/ src/
RUN . /opt/ros/humble/setup.sh && \
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release --merge-install
FROM ros:humble-ros-core AS runtime
COPY --from=builder /opt/ros2_ws/install /opt/ros2_ws/install
# Remove shell and package manager — prevents interactive exploitation
RUN rm -f /bin/sh /bin/bash /bin/dash && apt-get purge -y --auto-remove apttrivy image --severity HIGH,CRITICAL my-robot/navigation:latest
cosign sign --key cosign.key my-registry.io/my-robot/navigation:v1.2.3
cosign verify --key cosign.pub my-registry.io/my-robot/navigation:v1.2.3 || exit 1# docker-compose.yml
services:
motor_controller:
image: my-robot/motor-controller:v1.0.0
user: "1000:1000"
read_only: true
tmpfs: ["/tmp:size=64M", "/var/log/ros:size=32M"]
volumes:
- type: bind
source: /opt/robot/sros2_keystore/enclaves/my_robot/motor_controller
target: /keystore
read_only: true
security_opt: ["no-new-privileges:true"]
cap_drop: [ALL]
environment:
ROS_SECURITY_KEYSTORE: /keystore
ROS_SECURITY_ENABLE: "true"
ROS_SECURITY_STRATEGY: EnforceCyber attacks on robots cause physical harm. Standard IT security is necessary but not sufficient.
The emergency stop must function with all software, network, and main compute completely dead.
┌──────────┐ HARDWIRED ┌─────────────────┐
│ Physical │ ─────────────────▶│ Safety Relay / │──▶ Motor power cut
│ E-Stop │ Direct circuit │ Safety PLC │ via contactor
│ Button │ NO software └─────────────────┘
└──────────┘
┌──────────┐ OPTIONAL
│ Software │ ───(notifies)───▶ Can trigger relay, but NOT sole path
│ E-Stop │
└──────────┘
Main compute crash ──X──▶ Cannot prevent hardware e-stop
Network failure ──X──▶ Cannot prevent hardware e-stopDesign rules: hardwired circuit disconnects motor power; software triggers the relay but is never the only path; wireless e-stops use dedicated radio, not WiFi.
┌──────────────────────────────┬───────────────────────────────┐
│ MAIN COMPUTE (Jetson/x86) │ SAFETY CONTROLLER (STM32/MCU) │
│ Ubuntu + ROS2 │ Bare-metal firmware │
│ Nav, Perception, Planning │ │
│ ──── CAN/UART ──▶│ Validates: │
│ cmd_vel │ - Max velocity │
│ │ - Max acceleration │
│ ◀── joint_fb ────│ - Workspace limits │
│ │ - Watchdog timeout │
│ If compromised, safety │ Rejects out-of-bounds cmds │
│ controller STILL enforces │ Runs on separate hardware │
│ physical limits. │ Does NOT run ROS2 or Linux │
└──────────────────────────────┴───────────────────────────────┘Enforce at the driver level — last line of defense before actuators:
# velocity_safety_gate.py
import rclpy
from rclpy.node import Node
from geometry_msgs.msg import Twist
class VelocitySafetyGate(Node):
def __init__(self):
super().__init__('velocity_safety_gate')
self.declare_parameter('max_linear_vel', 1.0) # m/s
self.declare_parameter('max_angular_vel', 2.0) # rad/s
self.declare_parameter('max_linear_accel', 0.5) # m/s^2
self.declare_parameter('cmd_timeout_sec', 0.5)
self.declare_parameter('max_cmd_rate_hz', 50.0)
self.max_lin = self.get_parameter('max_linear_vel').value
self.max_ang = self.get_parameter('max_angular_vel').value
self.max_acc = self.get_parameter('max_linear_accel').value
self.timeout = self.get_parameter('cmd_timeout_sec').value
self.min_period = 1.0 / self.get_parameter('max_cmd_rate_hz').value
self.last_cmd_time = self.get_clock().now()
self.last_linear = 0.0
self.last_pub_sec = 0.0
self.sub = self.create_subscription(Twist, 'cmd_vel_raw', self.on_cmd, 10)
self.pub = self.create_publisher(Twist, 'cmd_vel', 10)
self.create_timer(0.1, self.watchdog_check)
def on_cmd(self, msg: Twist):
now = self.get_clock().now()
now_sec = now.nanoseconds / 1e9
if (now_sec - self.last_pub_sec) < self.min_period:
return # Rate limit exceeded — drop
msg.linear.x = max(-self.max_lin, min(self.max_lin, msg.linear.x))
msg.angular.z = max(-self.max_ang, min(self.max_ang, msg.angular.z))
dt = (now - self.last_cmd_time).nanoseconds / 1e9
if dt > 0:
accel = abs(msg.linear.x - self.last_linear) / dt
if accel > self.max_acc:
sign = 1.0 if msg.linear.x > self.last_linear else -1.0
msg.linear.x = self.last_linear + sign * self.max_acc * dt
self.pub.publish(msg)
self.last_cmd_time = now
self.last_linear = msg.linear.x
self.last_pub_sec = now_sec
def watchdog_check(self):
elapsed = (self.get_clock().now() - self.last_cmd_time).nanoseconds / 1e9
if elapsed > self.timeout:
self.pub.publish(Twist()) # No command → zero velocity# Hardware watchdog — kernel resets system if not fed
import os
class HardwareWatchdog:
"""Uses /dev/watchdog. If not fed within timeout, kernel triggers reset."""
def __init__(self):
self.fd = os.open('/dev/watchdog', os.O_WRONLY) # Starts countdown
def feed(self):
os.write(self.fd, b'\x00') # Reset countdown
def close(self):
os.write(self.fd, b'V') # Magic close — disarm gracefully
os.close(self.fd)# /etc/watchdog.conf
watchdog-device = /dev/watchdog
watchdog-timeout = 15
interval = 5
pidfile = /var/run/robot-safety-monitor.pid
max-load-1 = 24# /etc/fstab
/dev/mmcblk0p2 / ext4 ro,noatime,errors=remount-ro 0 1
tmpfs /tmp tmpfs nosuid,nodev,size=128M 0 0
tmpfs /var/log tmpfs nosuid,nodev,size=128M 0 0
/dev/mmcblk0p3 /data ext4 rw,noatime,nosuid,nodev 0 2# Alternative: overlayroot — all writes go to tmpfs, lost on reboot
sudo apt install overlayroot -y
# /etc/overlayroot.conf → overlayroot="tmpfs:swap=1"#!/bin/bash
set -euo pipefail
IMAGE="registry.myrobot.io/robot/navigation"
TAG="v$(cat VERSION)-$(git rev-parse --short HEAD)"
docker build -t "${IMAGE}:${TAG}" -f Dockerfile.prod .
trivy image --exit-code 1 --severity CRITICAL "${IMAGE}:${TAG}"
docker push "${IMAGE}:${TAG}"
cosign sign --key env://COSIGN_PRIVATE_KEY "${IMAGE}:${TAG}"
syft "${IMAGE}:${TAG}" -o spdx-json > sbom.json
cosign attach sbom --sbom sbom.json "${IMAGE}:${TAG}"# LUKS + TPM2 for unattended encrypted boot
sudo cryptsetup luksFormat /dev/mmcblk0p3
sudo cryptsetup luksOpen /dev/mmcblk0p3 robot-data
sudo systemd-cryptenroll /dev/mmcblk0p3 --tpm2-device=auto --tpm2-pcrs=0+7
# Disk decrypts only on original hardware with unmodified firmwarefrom pathlib import Path
from cryptography.hazmat.primitives.asymmetric import ec
from cryptography.hazmat.primitives import hashes, serialization
import subprocess
def verify_and_flash(firmware: Path, signature: Path, pubkey_path: Path):
pubkey = serialization.load_pem_public_key(pubkey_path.read_bytes())
try:
pubkey.verify(signature.read_bytes(), firmware.read_bytes(),
ec.ECDSA(hashes.SHA256()))
except Exception:
raise SecurityError("Firmware signature verification FAILED — aborting")
subprocess.run(['flash-tool', '--write', str(firmware)], check=True)# topic_auditor.py — logs publications on sensitive topics
import rclpy, json, time
from rclpy.node import Node
from geometry_msgs.msg import Twist
class TopicAuditor(Node):
def __init__(self):
super().__init__('topic_auditor')
self.log = open('/var/log/ros2_audit.jsonl', 'a')
self.create_subscription(Twist, '/cmd_vel', self.audit_cmd_vel, 10)
def audit_cmd_vel(self, msg: Twist):
record = {'ts': time.time(), 'topic': '/cmd_vel',
'lin_x': msg.linear.x, 'ang_z': msg.angular.z}
self.log.write(json.dumps(record) + '\n')
self.log.flush()
if abs(msg.linear.x) > 0.8 or abs(msg.angular.z) > 1.5:
self.get_logger().warn(f'HIGH VEL: lin={msg.linear.x:.2f} ang={msg.angular.z:.2f}')# cmd_vel_anomaly_detector.py
import numpy as np
from collections import deque
import rclpy
from rclpy.node import Node
from geometry_msgs.msg import Twist
class CmdVelAnomalyDetector(Node):
def __init__(self):
super().__init__('cmd_vel_anomaly_detector')
self.window = deque(maxlen=100)
self.alert_pub = self.create_publisher(Twist, '/security/cmd_vel_alert', 10)
self.create_subscription(Twist, '/cmd_vel', self.on_cmd, 10)
def on_cmd(self, msg: Twist):
self.window.append((msg.linear.x, msg.angular.z))
if len(self.window) < 20:
return
vels = np.array(list(self.window))
z_scores = np.abs((np.array([msg.linear.x, msg.angular.z]) - vels.mean(0)) / (vels.std(0) + 1e-6))
if np.any(z_scores > 3.0):
self.get_logger().error(f'ANOMALY: lin={msg.linear.x:.3f} ang={msg.angular.z:.3f} z={z_scores}')
self.alert_pub.publish(msg)# /etc/audit/rules.d/robot-security.rules
-w /opt/robot/sros2_keystore/ -p rwxa -k robot_keystore
-w /etc/robot/ -p wa -k robot_config
-w /home/robot-admin/.ssh/ -p wa -k ssh_keys
-w /opt/robot/firmware/ -p rwxa -k firmware_access
-w /etc/systemd/system/robot- -p wa -k robot_services
-a always,exit -F arch=b64 -F euid=0 -S execve -k root_commands
-w /dev/bus/usb/ -p rwxa -k usb_access
-w /etc/netplan/ -p wa -k network_config
-w /etc/iptables/ -p wa -k firewall_config
-w /usr/bin/docker -p x -k docker_execsudo auditctl -R /etc/audit/rules.d/robot-security.rules
sudo systemctl enable --now auditd
sudo ausearch -k robot_keystore --start todayProblem: Default ROS2 lets any DDS participant publish to /cmd_vel. One command from any machine on the LAN moves the robot.
# BAD: anyone on the network can do this
ros2 topic pub /cmd_vel geometry_msgs/Twist "{linear: {x: 999.0}}"Fix: SROS2 with Enforce. Restrict /cmd_vel publish to authorized enclaves. Velocity safety gate as secondary check.
# GOOD: unauthorized publish rejected at DDS layer
export ROS_SECURITY_ENABLE=true
export ROS_SECURITY_STRATEGY=EnforceProblem: One key compromised = entire fleet compromised.
# BAD: same key for all robots
ssh-copy-id -i ~/.ssh/fleet_key.pub robot@robot-001
ssh-copy-id -i ~/.ssh/fleet_key.pub robot@robot-002Fix: Unique key per robot. Use SSH CA with short-lived certificates.
# GOOD: SSH CA issues 8-hour certs per session
ssh-keygen -s /etc/ssh/ca_key -I "session-$(date +%s)" -n robot-admin -V +8h ~/.ssh/id_ed25519.pubProblem: Any compromised node = full root access to the system.
# BAD:
sudo ros2 launch my_robot bringup.launch.pyFix: Run as unprivileged user. Use udev rules for hardware access instead of root.
# GOOD:
sudo -u robot ros2 launch my_robot bringup.launch.py
# /etc/udev/rules.d/99-robot.rules:
# SUBSYSTEM=="tty", ATTRS{idVendor}=="0403", MODE="0660", GROUP="robot"Problem: All traffic on one flat network. Compromised IP camera reaches motor controller.
Fix: VLAN segmentation with inter-VLAN firewall rules. See Network Hardening section.
# BAD: everything on 192.168.1.0/24
# GOOD: VLAN 10 control (wired), VLAN 20 data, VLAN 30 mgmt (jump host)Problem: Credentials in version control exposed to repo access, CI logs, Docker layers.
# BAD: in params.yaml tracked by git
cloud_connector:
ros__parameters:
aws_access_key: "AKIAIOSFODNN7EXAMPLE"Fix: Environment variables from protected files. Scan repos with gitleaks.
# GOOD: secrets injected at runtime via systemd EnvironmentFile
gitleaks detect --source . --verbose # Pre-commit checkProblem: Software e-stop over ROS2 as the only safety mechanism. Network down = no stop.
# BAD: sole e-stop is a ROS2 topic subscriber
self.create_subscription(Bool, '/e_stop', self.software_estop, 10)Fix: Hardwired e-stop circuit. Software e-stop is an additional layer, never the sole path.
Problem: SROS2 certs generated once and never rotated. Compromised key = permanent access.
Fix: Monthly automated rotation via cron. Explicit validity periods in permissions XML. Emergency rotation capability via fleet management.
Problem: SROS2 disabled in production because "too hard" or "adds latency." Most common robotics security failure.
# BAD: "temporary" becomes permanent
export ROS_SECURITY_ENABLE=falseFix: Security enabled in CI/CD from day one. Tests must pass with Enforce.
# GOOD: CI enforces security
export ROS_SECURITY_ENABLE=true
export ROS_SECURITY_STRATEGY=Enforce
ros2 launch my_robot test.launch.py # Must pass with security onEnforce strategy — all nodes use encrypted, authenticated DDS© arpitg1304, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Just SKILL.md in skills/robotics-security of arpitg1304/robotics-agent-skills.
Open the folder on GitHubat commit f9bc546
Robotics Security next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Robotics Security this skillarpitg1304/robotics-agent-skills | 368 | — | ~7.8k | Automated safety check: Warn | Apache-2.0 | |
| Security AuditaAAaqwq/AGI-Super-Team | 105 | 2 repos | ~619 | Automated safety check: Notes | MIT | |
| Vercel Security Basicsjeremylongshore/tons-of-skills-marketplace | 2.8k | — | ~1.9k | Automated safety check: Notes | MIT | |
| Azure Devtest LabsMicrosoftDocs/Agent-Skills | 777 | — | ~3.5k | Automated safety check: Pass | CC-BY-4.0 | |
| Azure Key VaultKilo-Org/kilo-marketplace | 190 | 1 repos | ~1.9k | Automated safety check: Pass | MIT | |
| Security Hardeningchmonitor/chmonitor | 299 | — | ~440 | Automated safety check: Pass | GPL-3.0 |
aAAaqwq/AGI-Super-Team
Comprehensive security auditing for Clawdbot deployments. An agent skill from aAAaqwq/AGI-Super-Team.
jeremylongshore/tons-of-skills-marketplace
Apply Vercel security best practices for secrets, headers, and access control.
MicrosoftDocs/Agent-Skills
Expert knowledge for Azure DevTest Labs development including troubleshooting, best practices, decision making, architecture & design patterns, limits & quotas, security, configuration, integrations…
Kilo-Org/kilo-marketplace
Guidance for Azure Key Vault — securely storing and managing secrets, keys, and certificates with RBAC, network isolation, managed identity access, soft delete / purge protection, and rotation.
chmonitor/chmonitor
RBAC configuration, row policies, quotas, network security, audit logging, and access control best practices.
jeremylongshore/tons-of-skills-marketplace
Secure CoreWeave deployments with RBAC, network policies, and secrets management.
arpitg1304/robotics-agent-skills
Testing strategies, patterns, and tools for robotics software.
arpitg1304/robotics-agent-skills
Best practices, design patterns, and common pitfalls for ROS1 (Robot Operating System 1) development.
arpitg1304/robotics-agent-skills
Best practices for Docker-based ROS2 development including multi-stage Dockerfiles, docker-compose for multi-container robotic systems, DDS discovery across containers, GPU passthrough for…
arpitg1304/robotics-agent-skills
Bringing up a complete ROS2 system on a robot's onboard computer: systemd services, launch file composition, ordered startup, and production monitoring.
arpitg1304/robotics-agent-skills
Comprehensive best practices for robot perception systems covering cameras, LiDARs, depth sensors, IMUs, and multi-sensor setups.
arpitg1304/robotics-agent-skills
Architecture patterns, design principles, and proven recipes for building robust robotics software.
Categories
Security hardening and best practices for robotic systems, covering SROS2 DDS security, network segmentation, secrets management, secure boot, and the physical-cyber safety intersection. Robotics Security is an agent skill from arpitg1304/robotics-agent-skills. Security hardening and best practices for robotic systems, covering SROS2 DDS security, network segmentation, secrets management, secure boot, and the physical-cyber safety intersection.
Robotics Security fits situations like: securing ROS2 communications; configuring DDS encryption and access control; hardening robot onboard computers; managing certificates and credentials.
Run `npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a claude-code`. Or copy the skill folder (skills/robotics-security in arpitg1304/robotics-agent-skills) into .claude/skills/robotics-security in your project. Claude Code loads it when a task matches its description.
Run `npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a codex`. Or copy the skill folder (skills/robotics-security in arpitg1304/robotics-agent-skills) into .agents/skills/robotics-security in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add arpitg1304/robotics-agent-skills --skill robotics-security -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/robotics-security, .gemini/skills/robotics-security, .github/skills/robotics-security and .opencode/skills/robotics-security in your project.
Going by SKILL.md and its folder, Robotics Security needs the command-line tools its instructions call (apt, trivy, docker, git and gitleaks) and credentials named FLEET_API_KEY, AWS_SECRET_KEY and COSIGN_PRIVATE_KEY. Our summary lists: Python 3; Docker.
SKILL.md names 1 domain. In commands or code: w3.org; the agent is likely to contact it when it follows the instructions. This is read from the text; nothing was executed.
Our automated static check of SKILL.md flagged 2 warning(s): mentions a credentials file (ssh keys, cloud or package-manager tokens). Read the flagged lines before installing; the check is not a guarantee either way.
Robotics Security is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 7.8k tokens (SKILL.md is roughly 31k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
Skills that share tags, products or a category with Robotics Security: Security Audit (aAAaqwq/AGI-Super-Team, 105 stars), Vercel Security Basics (jeremylongshore/tons-of-skills-marketplace, 2.8k stars), Azure Devtest Labs (MicrosoftDocs/Agent-Skills, 777 stars) and Azure Key Vault (Kilo-Org/kilo-marketplace, 190 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
arpitg1304 (a GitHub user) maintains it in arpitg1304/robotics-agent-skills, which has 368 GitHub stars. The repository holds 10 skills in this directory. The repository was last updated on August 12, 2026.
Source: arpitg1304/robotics-agent-skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.