The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the cube CLI tool to perform operations such as schema discovery, data querying, and sending feedback, involving direct interaction with the underlying system.
[DATA_EXFILTRATION]: The cube feedback command is designed to send an email to the UNISIS support team containing the user's message and their identity. This is a documented transmission path that requires active cloud authentication.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes output from external SQLite databases and tool responses.
Ingestion points: Data retrieved via the cube schema and cube query commands (File: SKILL.md).
Boundary markers: The instructions do not define delimiters to separate untrusted data from the agent's system instructions.
Capability inventory: The agent can query database content and send external communications via email (File: SKILL.md).
Sanitization: No validation or sanitization of the data retrieved from the database is described.

cube