formative-assessment-loop-designer

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill uses a design that integrates untrusted user input into its system prompt instructions, creating a vulnerability surface for indirect prompt injection. However, the overall risk is minimal as the agent has no dangerous capabilities.
  • Ingestion points: User-provided inputs such as learning_objective and current_assessment_approach are interpolated directly into the model's instructions in SKILL.md.
  • Boundary markers: The prompt template does not utilize delimiters (e.g., triple quotes or XML tags) or specific isolation instructions to prevent the model from following commands contained within user input.
  • Capability inventory: Analysis of the skill reveals no access to sensitive tools, file system operations, network communication, or arbitrary command execution.
  • Sanitization: There are no validation or sanitization mechanisms implemented to filter or escape the user-provided text before interpolation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 04:53 AM
Security Audit — agent-trust-hub — formative-assessment-loop-designer