zarr-python

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill includes instructions to install Python packages such as zarr, s3fs, and gcsfs using the uv package manager.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch metadata from ArXiv (arxiv.org) to generate accurate citations. ArXiv is a well-known scientific repository and is treated as a safe source.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a process where the agent fetches and parses external data from ArXiv to construct a citation, which represents a potential ingestion point for untrusted data.
  • Ingestion points: The "Citing Scientific Agent Skills" section in SKILL.md instructs the agent to fetch external records from arxiv.org.
  • Boundary markers: No specific delimiters or "ignore instructions" warnings are provided for the fetched metadata.
  • Capability inventory: The skill allows access to Bash, Read, Write, and Edit tools.
  • Sanitization: The instructions do not specify validation or sanitization of the fetched citation strings before they are incorporated into the agent's output.
  • [SAFE]: The documentation provides explicit security best practices regarding cloud storage credentials, advising users to prefer IAM roles or workload identity and warning against printing or logging credential values.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 02:26 PM
Security Audit — agent-trust-hub — zarr-python