exploratory-data-analysis
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- Category 8: Indirect Prompt Injection (MEDIUM): The template populates report fields using untrusted data from analyzed files without sanitization or boundary markers. If an agent later reads this report, malicious instructions embedded in the analyzed data could hijack the agent's logic.
- Ingestion points: Multiple placeholders in assets/report_template.md including {FORMAT_DESCRIPTION}, {PATTERNS_FINDING}, and {COMPLETE_METADATA}.
- Boundary markers: Absent in the template structure.
- Capability inventory: Inferred file reading and analysis capabilities based on the template's purpose.
- Sanitization: Absent.
- Category 2: Data Exposure (LOW): The template explicitly outputs the absolute {FILEPATH} and {COMPLETE_METADATA}, which can leak sensitive system architecture or configuration details if the skill is used to analyze unauthorized files.
Audit Metadata