expense-categorization
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill inherently processes external, untrusted data in the form of transaction lists and receipts, which creates a theoretical surface for indirect prompt injection. 1. Ingestion points: Data enters the agent's context through bank statement imports and receipt processing (SKILL.md). 2. Boundary markers: The skill implements structured decision-making through references/DECISION-RULES.md and confidence scoring logic to determine when manual review is required. 3. Capability inventory: The execution environment allows the use of Read, Write, Edit, and WebFetch tools. A local Python script scripts/category_review_summary.py is used to aggregate data. 4. Sanitization: Risk is mitigated by requiring user review for all low-confidence mappings and ambiguous vendors, ensuring data does not automatically trigger downstream actions without oversight.
Audit Metadata