test-t3-mobile
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The script
pair-client.shexecutes a command string passed as an argument ($agent_device_command). This command is derived from theagentDevice.commandoutput provided by the platform's device tools. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external tool outputs without explicit sanitization before passing it to shell scripts.
- Ingestion points: Data enters the context via the
device_listandquickStartoutputs inSKILL.md. - Boundary markers: No explicit boundary markers or sanitization instructions are present for the ingested tool outputs.
- Capability inventory: The skill can execute shell scripts (
pair-client.sh), perform file system operations on SQLite fixtures, and interact with local network services. - Sanitization: There is no evidence of sanitization or escaping of the tool outputs before they are used in shell commands.
- [DYNAMIC_EXECUTION]: The
pair-client.shscript dynamically generates a deep link URL using an inline Node.js heredoc to process environment variables at runtime.
Audit Metadata