syllabus-evidence-based
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates entirely within the scope of document analysis and instructional design. It processes user-provided content (PDFs, images, text) and performs web searches to find comparable academic syllabi. No evidence of prompt injection, unauthorized data access, or malicious code execution was found.
- [EXTERNAL_DOWNLOADS]: The skill includes a research protocol (Step 1.5 and
references/syllabus-research-protocol.md) that directs the agent to perform web searches for existing syllabi from reputable academic sources (e.g., law school domains like .edu). These downloads are for informational purposes to inform syllabus structure and are considered safe under the trusted-scope rules for well-known educational institutions. - [COMMAND_EXECUTION]: The skill does not contain any shell commands or script execution logic. It relies on standard LLM capabilities for text analysis and search.
- [DATA_EXFILTRATION]: There are no patterns indicating that user-provided material is sent to unauthorized external domains. The research protocol is used to fetch public data into the context, not to export private user data.
- [PROMPT_INJECTION]: The instructions are focused on pedagogical structure. While it includes strong directives for the agent (e.g., 'You must not assign cases... without disclosing'), these are safety and quality controls rather than attempts to bypass model guardrails or extract system prompts.
Audit Metadata