service-omni-channel-setup-coordinate
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements a production environment guard ("safe_to_write") in "scripts/integration-driver.sh" that validates organization type and metadata before allowing any configuration changes.\n- [SAFE]: Secure credential management is enforced via the "redact_credentials" function, which moves passwords to a restricted file with "0600" permissions and redacts them from logs and artifacts.\n- [SAFE]: Input validation using regular expressions is applied to parameters like organization aliases and developer names to prevent command injection into the underlying shell and Salesforce CLI tools.\n- [COMMAND_EXECUTION]: The driver script orchestrates multiple internal child scripts and Salesforce CLI commands. This execution is governed by structural validation and production environment checks.\n- [INDIRECT_PROMPT_INJECTION]: The tool retrieves metadata from Salesforce organizations and interpolates it into markdown reports reviewed by the agent, representing an indirect injection surface.\n
- Ingestion points: Metadata such as Profile, Group, and Queue names retrieved via "sf data query" commands.\n
- Boundary markers: External data is included in the final report without specific structural delimiters or isolation markers.\n
- Capability inventory: The skill possesses capabilities for file modification and shell command execution.\n
- Sanitization: Input parameters are validated for shell safety, but the content of retrieved metadata is not sanitized for natural language instructions.
Audit Metadata