aer-consistency

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for internal consistency auditing of academic manuscripts. It provides a structured manual checklist and an automated script for verifying LaTeX and BibTeX integrity.
  • [COMMAND_EXECUTION]: The skill includes a Python script (scripts/audit_manuscript.py) intended to be run locally by the agent. The script uses only standard Python libraries (argparse, csv, re, sys, pathlib) and does not perform network operations or execute arbitrary shell commands.
  • [DATA_EXPOSURE_EXFILTRATION]: The automated script reads user-provided .tex, .bib, and .csv files to extract metadata (citations, labels, word counts). It includes security measures to prevent directory traversal when checking file references in the claim ledger by ensuring paths remain within the designated evidence root.
  • [INDIRECT_PROMPT_INJECTION]: While the skill processes external data (manuscript files), it performs deterministic parsing using regular expressions for specific structural elements (citations, labels, abstract blocks). It does not interpret or execute the content of the files as instructions for the agent.
  • [EXTERNAL_DOWNLOADS]: The skill does not perform any external downloads or remote code execution. All logic is contained within the provided scripts and instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 09:20 AM
Security Audit — agent-trust-hub — aer-consistency