literature-engineer

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill performs network requests to retrieve academic metadata and abstracts from export.arxiv.org and api.semanticscholar.org.
  • It utilizes the r.jina.ai proxy service to facilitate stable network connectivity in restricted environments.
  • All network interactions target well-known academic repositories or utility services for the primary purpose of literature retrieval.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted content from the web.
  • Ingestion points: Data is ingested through external API calls to ArXiv and Semantic Scholar, as well as local file imports (BibTeX, CSV, JSON) via scripts/run.py.
  • Boundary markers: The skill writes the collected metadata directly to papers/papers_raw.jsonl without applying boundary delimiters or adding warnings to ignore embedded instructions.
  • Capability inventory: The skill has the capability to perform network GET requests and write to the local filesystem.
  • Sanitization: Paper abstracts and titles are normalized for formatting but are not sanitized or filtered for potential prompt injection strings that could influence subsequent agent processing steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 01:59 PM