validation
Warn
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides the
setValidationCommandandstartValidationRunfunctions, which allow for the registration and execution of arbitrary shell commands. This capability, while central to the skill's purpose, enables the execution of any system command, posing a significant security risk if the agent is misused to perform unauthorized operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection through therunSummaryfield in thestartValidationRunoutput, which contains an LLM-generated summary of command logs.\n - Ingestion points: Command output logs located in
/tmp/validation/are read and processed by an LLM to generate summaries.\n - Boundary markers: There are no specified delimiters or instructions in the documentation to ensure the LLM ignores instructions that might be embedded in the log files.\n
- Capability inventory: The skill possesses the ability to register and execute arbitrary shell commands.\n
- Sanitization: The skill does not indicate that command output is sanitized or filtered for malicious content before being summarized by the LLM.
Audit Metadata