The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content such as syllabi and past papers, which constitutes an ingestion surface for potential instructions embedded in the data. However, the risk is mitigated by the agent's restricted toolset.\n
Ingestion points: Processes documents using Read, Glob, and Grep tools based on files provided by the user (e.g., in Full Roadmap Mode).\n
Boundary markers: The instructions do not explicitly define delimiters to separate user data from system instructions, nor do they include warnings to ignore embedded commands.\n
Capability inventory: The skill is restricted to Read, Glob, and Grep. It cannot perform network operations, write to the filesystem, or execute shell commands, preventing exfiltration or persistence.\n
Sanitization: No explicit sanitization or filtering of the uploaded study materials is performed before analysis.

examprep-ai