The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests data from external sources and project artifacts to generate lessons and assignments.
Ingestion points: The agent reads content from resources/source-files/ (externally sourced research), unit.md (instructional contracts), and Learner.md (conceptual mastery profiles).
Boundary markers: The instructions do not explicitly define specific delimiters for separating untrusted external research from internal agent logic, though they mandate distinguishing source claims from instructional synthesis.
Capability inventory: The agent has file-write access to create and update task.md (Markdown), assignment.html, and notes.html (HTML), and it updates the AGENTS.md ledger for state management.
Sanitization: While the agent is instructed to produce 'beautiful and purposeful' HTML, there are no specific technical requirements for sanitizing or escaping content derived from source-files before interpolation into lesson artifacts.

teacher