juicebox-migration-deep-dive

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requests administrative-level permissions in the metadata via Bash(kubectl:*) and Bash(curl:*). These tools are intended for performing bulk data imports and managing transformation pipelines in enterprise environments, though they represent a significant capability surface.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its primary function of ingesting and processing data from external sources such as LinkedIn Recruiter, Greenhouse, Lever, and CSV exports.
  • Ingestion points: Data entering via exports from LinkedIn Recruiter, Greenhouse, Lever, and local CSV/Excel files.
  • Boundary markers: Absent; there are no specific instructions to the agent to ignore or delimit embedded instructions within the source data.
  • Capability inventory: The agent has access to Read, Write, Edit, Bash(kubectl:*), and Bash(curl:*) permissions.
  • Sanitization: The provided code snippets perform structural validation and field mapping but do not include content sanitization to prevent the execution of instructions embedded in data fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 11:44 AM