gws-modelarmor-sanitize-response
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
gwscommand-line utility to interact with the Google Model Armor service. This is the intended functionality for sanitizing model outputs. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted model-generated content via the
--textor--jsonflags. While this creates a data ingestion surface, the skill's purpose is to mitigate risks by passing this data to a security sanitizer. - Ingestion points: The content for sanitization is passed through the
--textand--jsonflags in thegws modelarmor +sanitize-responsecommand. - Boundary markers: No specific delimiters are used in the example commands to isolate the untrusted text.
- Capability inventory: The skill performs CLI command execution via the
gwsbinary. - Sanitization: This skill functions as a sanitization control by utilizing Google Model Armor to filter content.
Audit Metadata