The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing the omniroute package globally via npm or executing it via npx. This package is an external dependency not hosted by a recognized trusted organization. \n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data that could contain malicious instructions. \n
Ingestion points: omniroute batches output, omniroute batches errors, and omniroute files content (in SKILL.md) allow the agent to read potentially untrusted data from the batch service. \n
Boundary markers: No explicit instruction delimiters or "ignore" warnings are provided in the skill instructions to separate retrieved data from agent instructions. \n
Capability inventory: The agent has capabilities to upload local files (omniroute files upload), write to the filesystem (via --out flags), and delete remote files (omniroute files delete). \n
Sanitization: There is no evidence of sanitization or filtering of the retrieved content before it enters the agent's context. \n- [DATA_EXFILTRATION]: The omniroute files upload <path> command enables the agent to send local files to a remote service. While this is the intended functionality of the CLI tool, it represents a risk if an agent is tricked into uploading sensitive configuration or credential files. \n- [COMMAND_EXECUTION]: The skill relies on executing the omniroute CLI tool for its operations, which involves executing shell commands via the agent's environment.

cli-batches