procurement-optimizer
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate utility for business operations. No patterns of prompt injection, data exfiltration, or malicious execution were found.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external spend inventory data provided in JSON format. While this creates a theoretical surface for indirect prompt injection (e.g., if tool names contained malicious instructions), the risk is minimal as the scripts perform only deterministic analysis and the data is intended for business reporting.\n
- Ingestion points: The scripts
license_utilization_analyzer.py,tool_overlap_detector.py, andsavings_opportunity_ranker.pyingest data from user-specified JSON files via the--inputargument.\n - Boundary markers: None explicitly defined in the prompts, though the scripts output structured text and JSON reports which help delineate data from instructions.\n
- Capability inventory: The skill uses Python scripts to perform file reads and data processing. No network operations, shell command execution of data, or writing to system files are present.\n
- Sanitization: The scripts parse the input as JSON and cast values to specific types (int, float, str), which provides a layer of data validation, though they do not explicitly sanitize for natural language instructions embedded in strings.
Audit Metadata