gws-modelarmor-create-template

Pass

Audited by Gen Agent Trust Hub on Mar 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the gws command-line tool with subcommands like modelarmor +create-template to interact with Google Cloud Platform services. It processes user-supplied flags for project ID, location, and template configuration to construct and run these commands.\n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through its template creation logic.\n
  • Ingestion points: The --json flag accepts arbitrary JSON strings which are used to define the Model Armor template configuration.\n
  • Boundary markers: The skill does not employ explicit boundary markers or provide instructions to the model to ignore instructions that might be embedded within the provided JSON configuration.\n
  • Capability inventory: The skill uses the gws binary to perform administrative write operations on cloud security infrastructure.\n
  • Sanitization: There is no evidence in the skill definition of sanitization, schema validation, or escaping of the content provided via the --json flag before it is used in command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 20, 2026, 04:56 PM