ask
Warn
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the @onequery/cli package globally using npm install -g @onequery/cli. This dependency is fetched from the public NPM registry and does not originate from a verified trusted vendor or well-known service.
- [COMMAND_EXECUTION]: The skill relies on executing shell commands via the onequery CLI for authentication, context resolution, and data querying, including operations like onequery auth login and onequery auth import which handles credentials.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests and summarizes data returned from external SQL queries.
- Ingestion points: Data returned from onequery query exec and onequery query validate commands in SKILL.md and references/command-patterns.md.
- Boundary markers: The instructions do not define clear delimiters (e.g., XML tags) to isolate external query results from the agent's instructions.
- Capability inventory: The skill possesses significant capabilities, including the ability to execute shell commands, install global software packages, and access the network through the CLI tool.
- Sanitization: No technical sanitization or validation is applied to the content returned by queries before it is processed by the agent.
Audit Metadata