The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the ajv CLI tool to validate course.json files against a predefined schema. This is a standard utility for JSON validation and is used here for structural integrity checks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted course data (course.json) to perform pedagogical audits and propose fixes. This presents a potential surface where malicious instructions in the course content could attempt to influence the agent's audit findings or the code diffs proposed in the 'Propose fixes' step.
Ingestion points: Reads docs/instructional-design/courses/{slug}/course.json.
Boundary markers: Structural validation is performed via JSON Schema (ajv), but no semantic delimiters are specified for the textual content of the lessons.
Capability inventory: Includes file writing (audit reports), file modification (applying diffs), and CLI execution (ajv).
Sanitization: Employs structural schema validation; however, it lacks specific sanitization for natural language instructions embedded within the course fields.

course-audit