metacognitive-prompt-library

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill constructs its final prompt by interpolating several user-supplied fields directly into the instruction block, which is a standard pattern for template-based agents but introduces an indirect prompt injection surface.
  • Ingestion points: Data from the task_description, student_level, metacognitive_focus, subject_area, student_profiles, and task_phase input fields in SKILL.md are used to populate the model's instructions.
  • Boundary markers: The prompt template lacks robust delimiters (such as XML tags or unique string markers) and specific instructions to the model to ignore any command-like text within the user-provided inputs.
  • Capability inventory: The skill's functionality is limited to text generation and lacks access to sensitive system operations like file system modification, network requests, or shell execution, which limits the potential for harm.
  • Sanitization: There is no evidence of input validation or filtering to remove potentially malicious instructions from the user-provided text.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 10:54 PM