referral-program
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface by encouraging the agent to read local context from
.claude/product-marketing-context.md. 1. Ingestion point: .claude/product-marketing-context.md; 2. Boundary markers: None explicitly specified in instructions; 3. Capability inventory: Limited to mathematical modeling and textual advice; 4. Sanitization: Standard LLM guardrails apply as no explicit sanitization is defined for the context file. - [COMMAND_EXECUTION]: The skill references and includes a local Python script
scripts/referral_roi_calculator.py. A security audit of the script confirms it is a safe mathematical utility using only built-in Python modules (json, sys, collections, argparse) with no network or filesystem-writing capabilities.
Audit Metadata