israeli-paid-ads
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to run a local script,
scripts/cpc_calculator.py, to estimate campaign costs based on user inputs. The script is part of the skill package and utilizes the Python standard library exclusively for arithmetic and argument parsing.\n- [INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided values (e.g., vertical name, budget) and passes them as command-line arguments to a local script. This creates a surface for indirect injection which is well-mitigated by the script's implementation.\n - Ingestion points: User parameters for budget and business vertical described in
SKILL.md.\n - Boundary markers: None explicitly defined in the instructions for user inputs.\n
- Capability inventory: Execution of a bundled Python script via shell commands.\n
- Sanitization: The script
scripts/cpc_calculator.pyusesargparseto strictly enforce numeric types for budget and CPC values and limits the vertical to a specific list of predefined choices.
Audit Metadata