customer-success-manager

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The Python scripts (health_score_calculator.py, churn_risk_analyzer.py, and expansion_opportunity_scorer.py) are implemented using only the Python standard library, ensuring no reliance on external or unverified third-party packages.
  • [SAFE]: No network access, file system modifications, or unauthorized command executions were found in the script logic. The tools operate strictly as local data processors.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing external JSON data (e.g., customer names and sentiment) and reflecting it into generated reports.
  • Ingestion points: input_file parameter processed by all three analytical scripts.
  • Boundary markers: Absent; raw strings from input files are interpolated directly into text and JSON output formats.
  • Capability inventory: Reading local JSON data and writing analysis results to standard output.
  • Sanitization: None; input data is parsed as JSON, but string content is not sanitized or escaped before being included in the output report.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 06:58 AM