document-reader

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a document reader using standard Python libraries and the textract package. It operates locally on files provided by the user and no malicious behavior or data exfiltration was found.
  • [PROMPT_INJECTION]: The skill processes untrusted external content from various document formats (PDF, DOCX, XLSX, etc.) and archive types (ZIP, RAR, 7z), creating a surface for indirect prompt injection.
  • Ingestion points: Document and archive reading methods in scripts/document_reader.py which extract content from user-provided files.
  • Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions for the extracted text; it outputs content with simple headers.
  • Capability inventory: The skill is capable of reading local files and writing temporary files to /tmp for processing. It does not perform network operations.
  • Sanitization: The script uses errors='replace' during text decoding but does not perform content sanitization or filtering of the extracted text.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 07:45 AM
Security Audit — agent-trust-hub — document-reader