paper-evaluator

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to install dependencies using pip install pdfplumber pypdf python-docx. While these are well-known and legitimate libraries for handling PDF and Word documents, they involve the execution of external setup commands.
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs third-party Python packages from public registries. These packages are widely used for the skill's primary purpose of document analysis.
  • [REMOTE_CODE_EXECUTION]: The skill includes Python code snippets for document parsing. The agent is expected to execute this code locally to extract text from 'paper.pdf' and 'paper.docx' for analysis.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted content from external documents which are then analyzed by the AI.
  • Ingestion points: Text is extracted from user-provided files such as 'paper.pdf' or 'paper.docx' in Phase 1-2.
  • Boundary markers: There are no explicit delimiters or 'ignore' instructions implemented to prevent the AI from following commands that might be embedded within the academic papers being evaluated.
  • Capability inventory: The agent has the capability to read files from the local filesystem and execute Python code for extraction purposes.
  • Sanitization: The skill lacks mechanisms to sanitize or filter the extracted text before it is used as input for the subsequent evaluation phases.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 06:29 PM