The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user-supplied data (lecture slides, homework, past papers) as its primary knowledge source. This creates an indirect prompt injection surface where malicious instructions could be embedded in PDF or text files. However, the skill implements strong mitigation strategies, including strict instruction grounding, provenance labeling (green/yellow/warning tags), and citation requirements.
[COMMAND_EXECUTION]: The skill uses subprocess.run() in multiple utility scripts to perform internal tasks. For example, cheatsheet_render.py invokes local browsers in headless mode to generate PDFs, and agent_smoke.py drives agent interfaces for testing purposes. These executions are constrained to the skill's own components and standard system utilities.
[EXTERNAL_DOWNLOADS]: The samples/fetch.py script downloads sample course materials from well-known and trusted educational domains, specifically MIT OpenCourseWare and Open Yale Courses. These are considered safe sources and do not represent a security risk.
[SAFE]: The skill instructs the user or agent to install standard Python libraries like pypdfium2 and pypdf for PDF processing. These are well-established packages from official registries.

exam-cram-coach