job-preference-interview

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security issues detected. The skill is designed to facilitate a job preferences interview and maintain a markdown file (preferences.md) within a local workspace.
  • [COMMAND_EXECUTION]: The skill includes a shell script, evals/files/setup-interview-ws.sh, which is used exclusively by the evaluation harness to set up a temporary environment. The script performs benign operations (creating a directory and echoing its path) and is not executed during normal runtime for the user.
  • [DATA_EXFILTRATION]: No network activity or patterns related to data exfiltration were found. The skill operates entirely on local workspace data and direct user input.
  • [PROMPT_INJECTION]: The skill contains strong behavioral instructions to ensure a specific interview flow (e.g., asking one question at a time), but these are task-oriented constraints rather than attempts to bypass AI safety guardrails or system prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 12:03 PM