nsfc-ref-alignment

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill initiates network requests to 'api.crossref.org' and 'api.openalex.org' to perform DOI validation. These are established scholarly metadata providers and are treated as safe sources according to the trust guidelines.\n- [PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by ingesting untrusted text from user-provided LaTeX and BibTeX files, which is subsequently provided to the host AI for semantic analysis. Maliciously formatted content within these files could attempt to subvert the AI's reasoning.\n
  • Ingestion points: Document text and metadata are read through 'latex_scanner.py' and 'bib_utils.py'.\n
  • Boundary markers: Although the data is passed in a structured JSON format, the skill does not implement specific boundary markers or delimiters in the text fields to instruct the AI to ignore potential instructions embedded in the data.\n
  • Capability inventory: The skill's write operations are restricted to the '.nsfc-ref-alignment/' directory and the user-defined 'references/' folder. It does not include functionality for arbitrary command execution or unauthorized network communication for data exfiltration.\n
  • Sanitization: While the script strips LaTeX comments, no specific sanitization or filtering is performed on the extracted content to mitigate prompt injection risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 03:41 PM