xray-paper-skill

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection. It processes untrusted academic content from external sources such as local PDF/text files and research paper URLs which are then incorporated into the analysis context.\n
  • Ingestion points: The skill uses WebFetch and a Python-based extractor (scripts/xray_io.py) to read content from user-provided paths and URLs.\n
  • Boundary markers: There are no explicit delimiters or instruction-isolation markers to prevent instructions embedded within the paper text from being followed by the agent.\n
  • Capability inventory: The skill utilizes Read, Write, WebFetch, and Bash tools, creating a significant attack surface if an attacker embeds malicious instructions in a paper.\n
  • Sanitization: Content is cleaned for logical extraction purposes but is not sanitized for potential malicious instructions or prompts.\n- [COMMAND_EXECUTION]: The skill executes a local script via Bash using parameters derived from user input ($ARGUMENTS). While the paths are enclosed in double quotes in the instructions, this pattern constitutes a potential command injection surface if the input is maliciously crafted and the underlying shell environment allows expansion or command chaining.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 03:17 PM