ref-check
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches bibliographic records and metadata from official domains of established academic services, including INSPIRE-HEP, arXiv, CrossRef, zbMATH, ADS, PubMed, and DBLP. These downloads are integral to the skill's stated purpose of reference verification.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from external academic registries, which could potentially contain malicious instructions embedded within paper titles, abstracts, or other fields.
- Ingestion points: API responses and landing pages from bibliographic databases mentioned in the SKILL.md instructions (e.g., inspirehep.net, arxiv.org, crossref.org).
- Boundary markers: There are no explicit instructions for the agent to use delimiters or to disregard instructions contained within the fetched bibliographic content.
- Capability inventory: The skill instructs the agent to fetch content from the web and write to local bibliography files (.bib); it also refers to shell commands like grep and LaTeX build tools.
- Sanitization: The skill lacks explicit sanitization or filtering requirements for the text fetched from external registries.
Audit Metadata