desktop-package
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use standard macOS commands
lsandopen. Thelscommand is used to verify that files were successfully placed in the target directory, andopenis used to show the folder in the Finder GUI for the user. These commands are restricted to the locally created folder path within~/Desktop. - [DATA_EXFILTRATION]: The skill accesses the filesystem to copy session-related files to the Desktop. While this involves data movement, the instructions explicitly prohibit the inclusion of sensitive materials such as API keys, tokens, or credentials, reducing the risk of accidental exposure of secrets.
- [PROMPT_INJECTION]: The skill processes conversation history to generate summary documents, which presents an indirect injection surface. The instructions mitigate this risk by requiring the agent to distill the content into clean standalone documents rather than transcribing raw chat data, and by maintaining strict boundaries against secret exposure.
Audit Metadata