nature-citation

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with the Crossref REST API (api.crossref.org) to retrieve scholarly metadata. Crossref is a well-known, established service for the scientific community, and these requests are necessary for the skill's primary function of citation discovery.
  • [COMMAND_EXECUTION]: The workflow relies on executing a local Python script (nature_citation.py) to segment text, perform network searches, and write reference files. The script uses only standard Python libraries and operates within the parameters provided by the user.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data in the form of article titles and abstracts from the Crossref API, which creates a potential surface for indirect prompt injection.
  • Ingestion points: External metadata is fetched in scripts/nature_citation.py via the fetch_crossref and fetch_crossref_doi functions.
  • Boundary markers: The results are structured into segment-based reports and a citation browser, separating source text from search results.
  • Capability inventory: The script has the capability to perform network GET requests and write various citation formats (JSON, TSV, HTML, ENW, RIS, RDF) to the local disk.
  • Sanitization: The script proactively sanitizes external content using html.escape, xml.sax.saxutils.escape, and json.dumps before generating any review artifacts, significantly mitigating the risk of malicious payloads affecting the agent or the user's browser.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 02:29 AM