lecture-review
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions specify the execution of a bundled Python script, scripts/analyze_stats.py, to generate statistical metrics from lecture transcripts. The script relies entirely on standard Python libraries to perform text processing tasks such as counting marker words, calculating speech rates, and identifying structural anchors in presentation decks.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing untrusted external data in the form of raw lecture transcriptions and HTML presentation decks.\n
- Ingestion points: Untrusted text data is ingested via file read operations and processed by both the agent and the statistical script.\n
- Boundary markers: The skill explicitly defines a core principle (Principle 5) that instructs the agent to limit its review to expression and structure, specifically forbidding the evaluation of content quality or methodology.\n
- Capability inventory: The skill is limited to local file system operations (reading transcripts/decks and writing reports/profiles) and local script execution.\n
- Sanitization: The statistical script applies regular expression filters to extract specific metrics and clean data for analysis.
Audit Metadata