alphagenome-atlas-website-links

Pass

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script scripts/alphagenome_atlas_links.py downloads GENCODE genomic annotations in Feather format from a Google-owned cloud storage bucket (storage.googleapis.com). This data is used locally to resolve gene coordinates and splice junctions for variant analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection via the processing of external variant records.
  • Ingestion points: The table subcommand in scripts/alphagenome_atlas_links.py ingests untrusted data from JSON files provided via the --input argument.
  • Boundary markers: The script uses Markdown backticks and table formatting to delimit ingested data in its output, which helps the agent distinguish data from instructions.
  • Capability inventory: The script is a data processing utility that lacks dangerous capabilities such as arbitrary file-system writes, command execution based on input content, or unauthorized network exfiltration of the ingested data.
  • Sanitization: The script uses urllib.parse.quote to sanitize data interpolated into URLs, reducing the risk of malformed or malicious link generation.
  • [COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to use uv run to execute the local utility script. These operations are restricted to the provided Python script and are necessary for the skill's primary functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 24, 2026, 03:14 PM