pp-human-goat
Warn
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of a wide array of terminal commands via the
human-goat-pp-clitool to manage account profiles, hiring tasks, and financial tracking. - [DATA_EXFILTRATION]: Provides a
--deliver webhook:<url>feature that allows the output of any command—potentially containing private account data, Tasker details, or invoice history—to be sent to an external, arbitrary URL. - [CREDENTIALS_UNSAFE]: The skill documents the storage locations for sensitive secrets, including
~/.magic/api_keyandcredentials.toml. It also describes a mechanism for lifting active session and XSRF cookies from the user's local Chrome browser. - [EXTERNAL_DOWNLOADS]: Instructs the user or agent to install the necessary CLI and MCP binaries using
go installfrom the author's GitHub repository andnpxfrom the author's NPM package namespace. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests natural language descriptions for hiring humans (
hire), dispatching remote errands (dispatch), and making phone calls (call). - Boundary markers: The instructions do not specify any delimiters or safety boundaries for these natural language inputs.
- Capability inventory: The skill has significant real-world impact capabilities, including the ability to spend money (autonomous checkout), hire local labor, and initiate phone calls.
- Sanitization: No sanitization or validation logic is defined for the input strings provided to the task hiring commands.
- [NO_CODE]: The skill itself consists of markdown instructions and command templates, containing no internal scripts or binary payloads.
Audit Metadata