nature-citation

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches bibliographic metadata and publication records from the official Crossref API (api.crossref.org). This is an established service for academic metadata and is essential to the skill's function.
  • [COMMAND_EXECUTION]: Utilizes a local Python script (scripts/nature_citation.py) to segment manuscript text, perform metadata searches, and generate formatted export files (RIS, ENW, Zotero RDF). The script is restricted to these operations and does not invoke untrusted shell commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted manuscript text.
  • Ingestion points: Passages provided by the user via command-line arguments or text files to the discovery script.
  • Boundary markers: The workflow enforces a strict separation between "candidates" (unverified metadata) and "references" (verified support). A manual screening gate (--screened-selection) is required before any reference-manager files are produced.
  • Capability inventory: The skill has network capabilities (API calls), file-writing capabilities (metadata and export files), and text segmentation logic.
  • Sanitization: The Python script implements proper XML escaping for RDF generation and regex-based sanitization for RIS and ENW formats to prevent injection into reference managers.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 08:48 AM