The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates several user-controlled input fields directly into the prompt instructions, which creates a potential surface for indirect prompt injection attacks. An adversary could provide malicious instructions within these fields to attempt to override the intended behavior of the agent. This surface is not mitigated by sanitization, although the impact is constrained by the absence of executable tools or network capabilities.
Ingestion points: Fields including media_type, analysis_focus, specific_text, and student_profiles (found in SKILL.md).
Boundary markers: The input variables are separated from the instructions using bold markdown headers.
Capability inventory: No external tools, shell execution, or file system access capabilities are defined or utilized.
Sanitization: No validation or sanitization is performed on user-supplied inputs prior to prompt interpolation.

media-literacy-deconstruction-protocol