alterlab-mixed-methods
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides methodological frameworks and Python-based logic for integrating qualitative and quantitative data, which involves processing research datasets. The capability to execute Python code via the
Bashtool and fetch external data viaWebFetchintroduces a potential attack surface if the agent processes data containing hidden malicious instructions. \n - Ingestion points: User-supplied research datasets (e.g., transcripts, survey results) and external web content retrieved via
WebFetch. \n - Boundary markers: Absent. The instructions do not define specific delimiters or "ignore" instructions for the agent when processing external data. \n
- Capability inventory: File access (
Read), network access (WebFetch,WebSearch), and Python execution (Bash). \n - Sanitization: The skill does not include logic to sanitize or validate the content of research data before it is processed by the agent or Python environment.\n- [DYNAMIC_EXECUTION]: The skill provides functional Python code snippets as templates for research tasks and explicitly supports the execution of Python scripts via the
Bashtool to facilitate integrated analysis. \n - Evidence:
SKILL.mdincludes theselect_qualitative_participantsPython function and references the use ofuv run pythonin thecompatibilityfield, which encourages the agent to generate and execute scripts at runtime to perform data sampling and integration.
Audit Metadata