gws-modelarmor-sanitize-prompt

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's functionality is consistent with its description as a prompt sanitization tool. It uses 'gws', a binary provided by the author 'googleworkspace', which is a trusted vendor resource.
  • [COMMAND_EXECUTION]: The skill executes the gws modelarmor +sanitize-prompt command. This is a legitimate use of a vendor-provided tool to perform the skill's primary function. All command arguments are handled according to standard CLI patterns.
  • [DATA_EXFILTRATION]: User input (provided via --text, --json, or stdin) is passed to the Google Model Armor service through the gws tool. This transmission is to the vendor's own infrastructure for the purpose of safety filtering and does not constitute unauthorized exfiltration.
  • [PROMPT_INJECTION]: No evidence of instructions designed to bypass safety filters or override agent behavior. The skill acts as a defense mechanism against prompt injection attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 10:49 PM