alterlab-scholar-eval
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill evaluates untrusted scholarly documents which could contain embedded instructions to influence the agent's scoring logic or behavior.
- Ingestion points: Research papers and theses read from the file system.
- Boundary markers: The instructions do not define specific delimiters for separating user-provided research text from agent instructions.
- Capability inventory: The agent uses Read/Write/Edit tools and can execute local scripts.
- Sanitization: No explicit sanitization of the research text is performed before dimension-based evaluation.
- [DYNAMIC_EXECUTION]: The skill invokes a local Python utility script to process quantitative data.
- Evidence:
SKILL.mdreferences the use ofscripts/calculate_scores.pyfor aggregate scoring. The script is a standard data processing tool included in the skill package that uses Python's standard library for JSON parsing and report generation.
Audit Metadata