research-paper-writing

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches LaTeX style definitions from the official NeurIPS domain (media.neurips.cc) and retrieves academic metadata from trusted providers including Semantic Scholar, CrossRef, and arXiv. These operations are essential for the skill's purpose and use reputable sources.
  • [COMMAND_EXECUTION]: Executes shell commands for LaTeX compilation, version control with Git, and background process management for long-running experiments. This includes the use of latexmk, pdflatex, and ps for legitimate monitoring.
  • [INDIRECT_PROMPT_INJECTION]: The pipeline ingests external text from academic sources via web tools. The risk of indirect injection is mitigated by structured verification workflows (cross-referencing citations) and explicit instructions to the agent to treat external content as unverified.
  • [DATA_EXFILTRATION]: No exfiltration of sensitive information or unauthorized network communication was detected. The skill includes specific guidance for the user to avoid including credentials or personal data in repositories.
  • [CREDENTIALS_UNSAFE]: No hardcoded API keys or secrets are present in the skill files. The instructions recommend standard practices for local secret management when using academic APIs.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:48 AM
Security Audit — agent-trust-hub — research-paper-writing