canvas-course-builder

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill correctly implements security best practices for secret management. It instructs the agent to read CANVAS_API_URL and CANVAS_API_TOKEN exclusively from environment variables, explicitly warning against hardcoding tokens in files or exposing them in user-facing output.
  • [COMMAND_EXECUTION]: The skill utilizes local Python scripts (scripts/build_course.py and scripts/canvas_client.py) to interface with the Canvas API. The scripts are designed for administrative automation and do not contain patterns for arbitrary or unsafe shell command execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes a capability to process external course specifications in YAML or JSON formats to build course content.
  • Ingestion points: External spec files loaded via the build_course.py script.
  • Boundary markers: None explicitly defined in the instructions, as the content is passed to the Canvas API.
  • Capability inventory: The skill can create, update, and manage course objects (assignments, modules, quizzes) via the Canvas REST API.
  • Sanitization: The skill relies on the Canvas LMS server-side HTML sanitization for all content fields, which is a standard and safe design for this integration.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 12:55 PM
Security Audit — agent-trust-hub — canvas-course-builder