agentforce-secure
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains numerous prompt injection patterns within its YAML payload files and documentation. These are legitimate test cases used to evaluate the security of other agents, not instructions designed to compromise the host agent. The evaluation logic for target agent responses is confined to classification, which prevents indirect injection from influencing the host.
- [EXTERNAL_DOWNLOADS]: The skill requires the pyyaml Python package and the Salesforce sf CLI. These are well-known tools from trusted sources and are essential for functionality.
- [COMMAND_EXECUTION]: The skill executes shell commands to interact with the Salesforce CLI and runs local Python scripts. These operations are necessary for retrieving agent configuration and executing security tests.
- [SAFE]: All identified adversarial patterns are contained within data assets or documentation meant for security testing, and the execution of scripts follows established development workflows for the target platform.
Audit Metadata