exampass
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches updates from its official repository on GitHub and installs standard packages from the Python Package Index (PyPI). These operations target well-known technology services and are intended for the primary purpose of maintaining the tool.
- [COMMAND_EXECUTION]: The
updatesubcommand inSKILL.mdexecutes a PowerShell script to perform repository synchronization tasks, such as stashing local changes and pulling from the master branch. This is a standard administrative function for user-requested updates. - [REMOTE_CODE_EXECUTION]: The update logic allows for the retrieval and execution of updated Python scripts from the author's public GitHub repository. This functionality is documented and restricted to a specific user-initiated command.
- [PROMPT_INJECTION]: The skill processes untrusted user-supplied documents (ingestion point:
run_exampass.py), which creates an indirect prompt injection surface. The system prompt templates in theagents/directory do not utilize explicit boundary markers for this data. However, the skill implements sanitization inscripts/template_engine.py(via_escape_stray_ltand_sanitize_questions) to prevent malicious content from executing in the generated HTML output. The agent's capabilities include local file system access and shell execution (for updates), but these are scoped to the intended functionality of a development and study tool.
Audit Metadata