differentiation-adapter

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill uses template variables to incorporate user-provided content directly into the prompt in SKILL.md, creating a potential surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the context through fields like original_task, learner_profile, and student_profiles.
  • Boundary markers: The prompt lacks explicit delimiters (such as triple quotes or XML tags) to separate these user-supplied inputs from the system's instructions.
  • Capability inventory: The skill is strictly informational and does not include any scripts or allowed-tools that would permit file system modifications, network exfiltration, or command execution.
  • Sanitization: There is no evidence of input validation or content filtering to prevent users from including adversarial instructions within the classroom task descriptions.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 04:53 AM