vs-search-tuning-specify-policy-direction
Warn
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions define workflows that involve executing shell commands using the
vs(Viking Search) CLI tool. User-supplied inputs, such as search queries (<q>) and request payloads (<requestPayload>), are interpolated directly into shell command templates (e.g.,vs search run --query "<q>"). This creates a significant surface for command injection if the agent does not strictly sanitize the input against shell metacharacters like semicolons, backticks, or pipes. - [CREDENTIALS_UNSAFE]: In section 2.4, the skill suggests troubleshooting authentication issues by passing Access Keys (AK) and Secret Keys (SK) explicitly using CLI flags (
--ak/--sk) or environment variables. Providing credentials as command-line arguments is a security risk because they can be exposed in the system's process list, shell history, or audit logs. - [EXTERNAL_DOWNLOADS]: The skill depends on the presence and specific version (
>=0.2.0) of thevsCLI tool. While this tool is part of the 'volcengine' ecosystem, it requires local installation and execution permissions to function, representing an external dependency for the agent's operations. - [DATA_EXFILTRATION]: The skill frequently interacts with the Viking Search service via the CLI to transmit and receive data, including
application-id,dataset-id, and search configurations. While these operations are consistent with the skill's stated purpose, the use of a CLI that performs network operations with sensitive identifiers should be monitored.
Audit Metadata