invoice-organizer

Pass

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data (text extracted from invoices and receipts) which could contain hidden instructions designed to manipulate the agent's behavior.\n
  • Ingestion points: Files identified by find and processed via text extraction from PDFs and images in SKILL.md.\n
  • Boundary markers: Absent. The skill does not instruct the agent to distinguish between invoice data and potential commands.\n
  • Capability inventory: File system manipulation (mkdir, cp, mv) and report generation in SKILL.md.\n
  • Sanitization: Absent. Extracted text is used directly in filenames and CSV reports without validation.\n- [DATA_EXFILTRATION]: The skill is designed to read and aggregate sensitive financial information (vendor names, amounts, tax details). Although it currently operates locally, this aggregation creates a consolidated source of sensitive data that could be targeted for exfiltration.\n- [COMMAND_EXECUTION]: The skill uses shell commands (find, mkdir, cp, mv) to manage files and includes an 'Automation Setup' section that directs the agent to generate persistent scripts for folder watching, which could be abused to maintain persistence or execute unauthorized code.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 18, 2026, 04:24 PM
Security Audit — agent-trust-hub — invoice-organizer