churn-prevention

Warn

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a local script that is not provided in the bundle, making its behavior unverifiable.\n
  • Evidence: The file SKILL.md contains a shell command python3 scripts/churn_impact_calculator.py.\n
  • Evidence: The "Output Artifacts" section explicitly directs the agent to "Run churn_impact_calculator.py with your inputs" when modeling churn impact.\n- [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection due to unvalidated data ingestion.\n
  • Ingestion points: The agent is instructed to read marketing-context.md (if present) before beginning its task, as noted in SKILL.md.\n
  • Boundary markers: There are no specified delimiters or instructions to treat the content of marketing-context.md as untrusted data.\n
  • Capability inventory: The agent is empowered to execute shell commands and perform business logic calculations.\n
  • Sanitization: No sanitization or validation logic is defined for the external context being ingested.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 5, 2026, 08:06 PM
Security Audit — agent-trust-hub — churn-prevention