integrity-forensics

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands to create directories, manage git repositories, and execute Python scripts during its bootstrap and execution phases.
  • [EXTERNAL_DOWNLOADS]: Fetches resources from the vendor's GitHub repository (wanshuiyin/Anti-Autoresearch) to set up the local auditing environment.
  • [REMOTE_CODE_EXECUTION]: Executes a Python script (eval/run_eval.py) and other workflows from the downloaded repository. The use of a fixed commit SHA (b47af6f983b38347b6d2110379e266400597cf66) mitigates the risk of unauthorized updates to the remote code.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of papers (PDFs, directories, or Arxiv IDs).
  • Ingestion points: Input passed via $ARGUMENTS representing the paper to be audited.
  • Boundary markers: The skill does not explicitly define delimiters or provide instructions to the agent to ignore potential instructions embedded within the paper content.
  • Capability inventory: The skill has the ability to execute shell commands, perform network operations (git clone), and run Python scripts.
  • Sanitization: No explicit sanitization of the input arguments or the content of the ingested papers is performed before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 09:36 AM