The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read and search through sensitive user configuration files including shell profiles (~/.bashrc, ~/.zshrc, ~~/.profile) and all environment files (~~/.env*) to locate exported variables. This exposes the content of files typically used for storing credentials and private environment configuration to the agent context.
[PRIVILEGE_ESCALATION]: The instructions include multiple commands requiring root privileges via sudo. Specifically, it directs the agent to modify system-level group memberships (usermod -aG docker), restart system services via systemctl, and manually manage the Docker daemon process with elevated privileges.
[PERSISTENCE]: In the troubleshooting section, the skill recommends establishing persistence by adding an @reboot entry to the crontab to automatically execute the Docker daemon on system startup.
[COMMAND_EXECUTION]: The skill relies on extensive use of the Bash tool to perform system introspection and configuration. It executes complex command chains to verify the presence of Docker, Compose, and NVIDIA GPU drivers, and manages infrastructure deployment via Helm and Docker Compose.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting data from untrusted local files.
Ingestion points: The agent is instructed to use grep on shell profiles (~~/.bashrc) and environment files (~~/.env*) in SKILL.md.
Boundary markers: No delimiters or instructions are provided to the agent to ignore or sanitize embedded instructions found within these files.
Capability inventory: The skill has access to the Bash tool, allowing for arbitrary subprocess execution, and manages Docker and Kubernetes environments.
Sanitization: There is no evidence of validation or escaping of the content read from the local files before it is processed by the agent.

deploy-usdsearch