box

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill reads file content from Box via the get_file_content tool. This creates a surface for indirect prompt injection where instructions hidden within Box documents could influence agent behavior.
  • Ingestion points: get_file_content method in SKILL.md.
  • Boundary markers: Absent. No delimiters or instructions are provided to the agent to treat file content as data only.
  • Capability inventory: File system writes (download), file deletion (delete-file), and network access to Box APIs.
  • Sanitization: Absent.
  • [COMMAND_EXECUTION]: The skill functions by executing a local command-line interface binary located at /opt/hatch/bin/box-cli. All Box operations (search, read, write) are routed through this tool.
  • [DATA_EXFILTRATION]: The skill facilitates data movement between the local workspace and Box.com. While this is the intended purpose, it includes capabilities to upload local files (upload) and download remote content (download), which requires user consent according to the manifest.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 01:17 AM