master-curriculum

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied parameters to generate customized learning paths, creating a potential surface for indirect prompt injection.\n
  • Ingestion points: User input collected for 'Target Tradition', 'Current Position' (L0-L3), and 'Reality Constraints' (e.g., time availability, language limits) in SKILL.md.\n
  • Boundary markers: The skill lacks explicit delimiters for user input but enforces a rigid markdown output template with mandatory headers ('根基', '深入', '精研', '可能的盲点') to constrain the agent's response.\n
  • Capability inventory: The skill can read internal repository files (references and other master skills) and generate links to related skills (/master-).\n
  • Sanitization: No explicit sanitization or validation logic is provided for the 'Reality Constraints' free-text input.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 01:52 PM