The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential attack surface where instructions embedded in source markdown files could influence the generated HTML content.
Ingestion points: Markdown files located in the local teaching directory /Users/vishal/teaching/badm350/modules/.
Boundary markers: The skill uses template placeholders such as {OBJECTIVES_LIST} and {SESSION1_TITLE} to interpolate content, but it lacks explicit safety delimiters or 'ignore embedded instructions' warnings for the agent.
Capability inventory: The skill employs Canvas API tools (mcp__canvas-api__create_page, mcp__canvas-api__create_assignment) to publish the resulting content to an LMS course.
Sanitization: There is no evidence of sanitization or validation of the ingested markdown content before it is converted to HTML.
[EXTERNAL_DOWNLOADS]: The generated HTML templates include links to external JavaScript, CSS, and image assets.
Assets are hosted on instructure-uploads.s3.amazonaws.com (official Canvas hosting) and files.ciditools.com (Cidi Labs design tools). These are recognized as well-known services within the educational technology ecosystem.
[DYNAMIC_EXECUTION]: The skill generates HTML templates containing remote script tags that execute in the context of the user's LMS environment.
Evidence: Both references/assignment-templates.md and references/start-here-template.md include <script src="https://instructure-uploads.s3.amazonaws.com/account_145590000000000001/attachments/9675810/dp_app.js"></script>.

badm350-canvas-builder