summarize
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
uvxto dynamically runmarkitdown[pdf]from the Python Package Index (PyPI). This is a standard way to use well-known document conversion utilities. - [COMMAND_EXECUTION]: The script
to-markdown.mjsexecutes shell commands viaspawnSyncto runmarkitdownand thepiCLI. These executions are constrained to the intended functionality of document processing and summarization. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (web pages and documents) and passes them to an LLM for summarization. While this presents an attack surface where a document could contain hidden instructions for the summarizer, the impact is low as the output is presented back to the user/agent for inspection, and the summarization uses a restricted 'no-tools' session.
Audit Metadata