interview-system-designer

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill includes three Python tools (hiring_calibrator.py, loop_designer.py, and question_bank_generator.py) that process data from external JSON files, creating a potential surface for indirect prompt injection attacks.
  • Ingestion points: Untrusted data enters the environment via the --input command-line argument used by all three Python scripts to load role definitions or interview results.
  • Boundary markers: The skill lacks specific delimiters or 'ignore' instructions that would prevent the agent from following malicious commands embedded within the input JSON data.
  • Capability inventory: The tools have the capability to write files to the local file system using the --output argument, which could be misused if an injection is successful.
  • Sanitization: The Python scripts parse the JSON files but do not perform sanitization or filtering to remove potential natural language instructions before including the data in generated reports and templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 12:32 PM
Security Audit — agent-trust-hub — interview-system-designer