The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill relies on reading local workspace files to establish the teaching context and 'zone of proximal development', which presents a surface for instructions embedded in those files to influence the agent.
Ingestion points: The agent reads from MISSION.md, RESOURCES.md, learning-records/*.md, and NOTES.md in the local workspace directory.
Boundary markers: None. The instructions do not specify using delimiters or cautioning the model about the contents of these files.
Capability inventory: The skill allows reading and writing files in the workspace and executing shell commands (e.g., to open files).
Sanitization: None. There is no mention of sanitizing or validating the data read from these files before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to 'open the lesson file for the user by running a CLI command', which involves the execution of shell commands using paths within the local directory.
[DYNAMIC_EXECUTION]: The agent is tasked with generating self-contained HTML lessons that include 'interactive' components like quizzes and simulators, which involve generating and saving executable web code (HTML/JS) to the local filesystem.

teach