interview

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Python utility scripts (interview_planner.py, build_sheet_builder.py, primitives_validator.py) to transform user interview responses into structured JSON artifacts. These scripts use standard libraries and do not perform network operations.
  • [INDIRECT_PROMPT_INJECTION]: The skill serves as an ingestion point for user-provided data (job descriptions, trigger events, and rubrics) which are then stored in a build-sheet.json file. While this creates an attack surface for indirect prompt injection in the resulting agent, the skill's own validator script (primitives_validator.py) explicitly detects and warns the user about these risks when read/write memory access is enabled.
  • Ingestion points: User input captured via AskUserQuestion and passed as arguments to interview_planner.py as seen in SKILL.md.
  • Boundary markers: None explicitly defined in the generated JSON, though the agent is instructed to capture words exactly.
  • Capability inventory: File system writes (JSON artifacts) and local script execution in scripts/interview_planner.py and scripts/build_sheet_builder.py.
  • Sanitization: Limited to string normalization (slugification, whitespace stripping) without escaping for prompt interpolation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 10:34 PM
Security Audit — agent-trust-hub — interview