memento-flashcards
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from YouTube transcripts and user-provided CSV files. This creates a surface for indirect prompt injection where malicious content in a transcript or CSV could attempt to influence the agent's behavior during quiz generation or answer grading.
- Ingestion points: Transcripts are fetched from YouTube via youtube_quiz.py; flashcards are imported from user-provided files via memento_cards.py import.
- Boundary markers: The prompt instructions for generating quiz questions from transcripts lack explicit boundary markers (such as XML tags or delimiters) to separate the transcript content from the instructions.
- Capability inventory: The skill has the capability to write to the file system (via memento_cards.py) and perform network operations (via youtube-transcript-api in youtube_quiz.py).
- Sanitization: No explicit sanitization or filtering of the transcript or CSV content is performed before processing.
- [EXTERNAL_DOWNLOADS]: The skill uses the youtube-transcript-api library to fetch content from YouTube. This is an expected dependency for the declared quiz generation feature and is managed through the platform's package management system.
Audit Metadata