paper-search

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses uv run to execute a command-line interface tool named paper-search. This allows the agent to interact with the local file system and invoke a Python-based utility to perform searches and file operations.
  • [EXTERNAL_DOWNLOADS]: The skill downloads PDF files and metadata from external scholarly databases including arXiv, PubMed, and others, saving them to the user's local directory.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface when processing academic papers. 1. Ingestion points: The read command extracts text from untrusted academic papers retrieved from external sources. 2. Boundary markers: No delimiters or ignore-instructions warnings are specified in the prompts to distinguish between paper content and agent instructions. 3. Capability inventory: The skill has the capability to execute shell commands via uv run and write files via the download command. 4. Sanitization: The skill does not provide instructions for sanitizing or filtering the extracted paper text before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 03:24 PM
Security Audit — agent-trust-hub — paper-search