The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute various shell commands as part of the intended workflow. This includes running the canvaslms CLI tool to fetch format examples, executing a local Python validation script (validate_quiz.py), and using make to push quiz data to the learning management system.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and analyze external quiz files (INL1Quiz-*.json) which represents a surface for indirect prompt injection if the content is malicious.
Ingestion points: The agent reads INL1Quiz-*.json files from the local file system (e.g., modules/week-N/) to perform quality reviews, redundancy analysis, and distractor balance checks.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to treat the ingested quiz content as untrusted data.
Capability inventory: The agent has the capability to execute shell commands (canvaslms, python3, make) which could be targeted by an injection attack.
Sanitization: The provided validate_quiz.py script performs structural validation of the JSON format and logical checks on scoring data, but it does not sanitize the text content for potential instruction-based injection patterns.

canvas-quiz