paper-siphon

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute the uvx paper-siphon command to process PDF files and convert them into Markdown formatted text.- [EXTERNAL_DOWNLOADS]: The skill uses uvx to fetch the paper-siphon package from a remote registry at runtime. It also performs network requests to download PDF content from external URLs provided by the user or agent.- [REMOTE_CODE_EXECUTION]: By using the uvx tool-runner, the skill downloads and executes the paper-siphon Python package from an external registry. This is a standard operation for this tool but involves running code fetched from a remote source.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from academic PDFs, which could contain adversarial instructions.
  • Ingestion points: Reads content from PDF files located at local paths or remote URLs.
  • Boundary markers: None are specified in the skill instructions to separate the extracted PDF content from the agent's system instructions.
  • Capability inventory: Execution of CLI commands, network access to fetch documents, and filesystem access to write output files.
  • Sanitization: The skill does not describe any specific sanitization, filtering, or validation of the extracted text before it is returned to the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 01:54 AM