motivation-diagnostic-task-redesign

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is entirely composed of natural language instructions and does not include any executable scripts, binaries, or code files.
  • [SAFE]: No network access, data exfiltration patterns, or hardcoded credentials were detected. The skill operates within the conversation context without external communication.
  • [PROMPT_INJECTION]: The skill processes untrusted input through fields like task_description and student_profiles. These inputs are interpolated into the prompt using standard labels (e.g., **Task description:** {{task_description}}) without the use of robust delimiters or specific instructions to ignore embedded instructions. This represents a surface for indirect prompt injection. However, because the skill has no tool-use capabilities, network access, or file system permissions, the risk is negligible and restricted to the manipulation of the generated text response.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 10:53 PM