cohort-analysis

Warn

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions specify that cohort analysis results should be saved via a script named churn-predictor.py. This indicates the execution of a local Python script to handle data persistence and longitudinal tracking.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the data it processes.
  • Ingestion points: The skill ingests data from external sources via CRM and Analytics Model Context Protocol (MCP) tools, including customer records, transaction histories, and engagement events.
  • Boundary markers: The instructions do not define boundary markers or provide explicit directions for the agent to ignore instructions embedded within the customer data.
  • Capability inventory: The skill possesses the ability to read and write files within the user's home directory (~/.claude-marketing/) and execute the churn-predictor.py script.
  • Sanitization: There are no apparent sanitization or validation steps for the data retrieved from MCP tools before it is used to generate reports or passed to the persistence script.
  • [DATA_EXPOSURE]: The skill is configured to access sensitive business information, including brand profiles, lifecycle definitions, and marketing guidelines, stored in the ~/.claude-marketing/brands/ directory.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 27, 2026, 01:02 PM