deploy-stack

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The guide instructs capturing and exporting sensitive stack outputs (e.g., DCVCredentials, Ecr login password flow) and echoing/using them in commands, which requires the agent to handle and potentially emit secret values verbatim, creating an exfiltration risk.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches and executes external images/code at runtime (e.g., pulls and runs the NVIDIA IsaacLab container nvcr.io/nvidia/isaac-lab:2.3.0, pulls/runs the training image from ECR like .dkr.ecr..amazonaws.com/gr00t-finetune:latest, clones/builds https://github.com/NVIDIA/Isaac-GR00T.git in the Dockerfile/CodeBuild, and downloads LeIsaac assets from https://github.com/LightwheelAI/leisaac/releases/tag/v0.1.0), so these URLs supply remote code that is executed and are required for the skill to run.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs privileged actions that modify system state—e.g., using sudo to install the Session Manager plugin, writing to /home/ubuntu/.ssh/authorized_keys via SSM (chown/chmod), and editing ~/.ssh/config and EC2 instance attributes and resources—thereby directing the agent to perform actions that change system files, enable remote access, and alter machine state.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)