ltv-cohort-read-ecommerce

Pass

Audited by Gen Agent Trust Hub on Nov 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data provided in CSV format, which represents a potential surface for indirect prompt injection.
  • Ingestion points: Order exports, margin CSVs, and COGS exports are read from the file system (referenced in SKILL.md).
  • Boundary markers: The instructions do not define specific delimiters for the data but suggest the agent should not read the export directly.
  • Capability inventory: The skill executes a local Python script ../scripts/cohort_read.py using system commands.
  • Sanitization: The skill instructs the agent to use a specialized script to generate tables rather than reading raw exports, which serves as a sanitization layer between the untrusted data and the model prompt.
  • [COMMAND_EXECUTION]: The workflow requires the agent to execute a local script, ../scripts/cohort_read.py, via the command line to process user-provided files. This is a legitimate functional requirement of the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 22, 2026, 01:24 AM