The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest information from external resources to create lessons and reference materials, creating a surface where malicious instructions in those resources could influence the agent's behavior.\n
Ingestion points: In SKILL.md, the agent is instructed to find high-quality resources to populate RESOURCES.md and then use those resources to acquire knowledge for lessons.\n
Boundary markers: The skill does not provide instructions to use delimiters or 'ignore embedded instructions' warnings when reading or processing content from these external resources.\n
Capability inventory: The skill can write files to the local workspace (e.g., ./lessons/, ./reference/) and invoke other skills like create-note to modify a user's Obsidian vault, as mentioned in the 'End Of A Lesson' section of SKILL.md.\n
Sanitization: There are no explicit instructions to sanitize or validate content retrieved from external sources before it is used to generate lesson content.\n- [DYNAMIC_EXECUTION]: The skill generates interactive HTML files which may contain JavaScript for quizzes and simulators, and encourages the user to open these files in their browser.\n
Evidence: SKILL.md and references/html-ui.md guide the agent to create self-contained HTML lessons in the ./lessons/ directory, specifically mentioning 'interactive lessons' using 'quizzes and light in-browser tasks' and 'simulators' that provide immediate feedback.\n
Risk: If the agent's content generation is compromised (e.g., via indirect prompt injection), it could write malicious scripts into these HTML files, which would then be executed in the user's browser context when opened for review as suggested by the skill instructions.

teach