research-paper-writing

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from academic papers and web sources during the literature review phase.
  • Ingestion points: SKILL.md (utilizes web_search and web_extract to retrieve paper abstracts and content from the internet).
  • Boundary markers: Absent; instructions do not explicitly require delimiters or warnings to ignore instructions that might be embedded in retrieved documents.
  • Capability inventory: The pipeline involves Python execution (execute_code) and terminal operations for processing data and compiling LaTeX.
  • Sanitization: Ingested paper content is processed without explicit filtering, though the context is restricted to scientific research.
  • [EXTERNAL_DOWNLOADS]: The skill contains automated processes for retrieving necessary research resources from official sources.
  • Fetches style files from the official NeurIPS conference domain (media.neurips.cc) via a Makefile build step.
  • References official conference repositories on GitHub (e.g., github.com/acl-org, github.com/ICLR) for template acquisition.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 05:58 AM