space-weread-export

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill orchestrates the execution of a Python script (scripts/md_to_pdf.py) to handle PDF conversion. This script invokes a headless browser (such as Chrome or Edge) via subprocess.run to render HTML content into a PDF document.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. It retrieves highlight and review data from the external WeRead API and embeds it into a document template without sufficient sanitization. In scripts/md_to_pdf.py, the naive_md function lacks HTML escaping, meaning any HTML tags (e.g., <script>, <iframe>) present in the user's book highlights will be rendered by the headless browser during the conversion process.\n
  • Ingestion points: Data entering via WeRead API endpoints /book/bookmarklist and /review/list/mine.\n
  • Boundary markers: No delimiters or instructions are used to prevent the interpreter from processing embedded commands in the note content.\n
  • Capability inventory: The skill can create directories, write files (Markdown and PDF), and execute subprocesses (Python and Browser).\n
  • Sanitization: HTML escaping is absent in the primary text processing path in scripts/md_to_pdf.py.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 01:38 AM
Security Audit — agent-trust-hub — space-weread-export