alterlab-qualitative-methods
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides Python scripts for automated coding and analysis of research transcripts, which constitutes a vulnerability surface for indirect instructions embedded in external data.
- Ingestion points: Text data is ingested from local transcript files (e.g.,
data/P01_transcript.txt) using standard file read operations. - Boundary markers: The provided code templates lack explicit boundary markers or instruction isolation techniques when processing transcript content.
- Capability inventory: The skill includes capabilities to read local files, execute text analysis logic, and save visualization outputs to the file system.
- Sanitization: The keyword-based coding logic does not perform sanitization of the input text to prevent potential instruction following.
- [EXTERNAL_DOWNLOADS]: The Python analysis templates include calls to
nltk.download('stopwords'), which fetches standard linguistic data from the official NLTK project repositories.
Audit Metadata