The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the user to create a sudoers file that grants passwordless root access to the current user using the command echo "$USER ALL=(ALL) NOPASSWD: ALL" | sudo tee /etc/sudoers.d/hermes-install. This bypasses the operating system's fundamental security boundary, allowing the user (and the agent) to execute commands with administrative privileges without verification.
[COMMAND_EXECUTION]: The skill provides tools for the agent to perform arbitrary shell command execution (/shell) and code execution (/code) on the local system, which can be exploited if the agent is manipulated by malicious input.
[REMOTE_CODE_EXECUTION]: The installation process involves downloading pre-built binaries and executing remote installation scripts from third-party GitHub repositories not listed as trusted sources.
[EXTERNAL_DOWNLOADS]: The skill directs users to download and install various executables (.dmg, .AppImage, .deb, .rpm, .exe) from external sources, bypassing platform security measures like macOS Gatekeeper and Windows SmartScreen via manual override instructions.
[INDIRECT_PROMPT_INJECTION]: The skill creates a wide attack surface by ingesting untrusted data from 16 different messaging platforms and web search results while having access to high-privilege system tools.
Ingestion points: Data enters the agent's context through integrated messaging gateways (Telegram, Discord, Slack, etc.) and web browsing tools.
Boundary markers: There are no explicit delimiters or instructions defined to isolate external data from the agent's core instruction set.
Capability inventory: The agent has the ability to write to the file system, execute terminal commands, and run Python code.
Sanitization: The skill lacks any description of input validation or sanitization mechanisms for the data received from external sources.

hermes-desktop-companion