ds-churn-signals

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data from Stripe records and manual uploads, creating an inherent surface for indirect prompt injection.
  • Ingestion points: External data enters the context through Stripe subscription details, cancellation reasons, and user-uploaded CSV or JSON files.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the reporting templates.
  • Capability inventory: The skill can read local files and execute shell commands via a utility script.
  • Sanitization: The skill relies on an external script for calculation, but the resulting data is interpreted by the model without specific sanitization of field contents.
  • [COMMAND_EXECUTION]: The skill uses shell commands to load business context and process data.
  • Evidence: The skill executes 'cat' to read product marketing context files and 'python' to run the 'ds_utils.py' script.
  • Context: These executions are restricted to vendor-provided paths and are necessary for the primary function of the churn analysis tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 03:40 AM