wordpress-woocommerce-development
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides PHP implementation examples for AI-powered features that are vulnerable to indirect prompt injection attacks. \n
- Ingestion points: Untrusted data enters the context through the
$questionvariable in the AI Shopping Assistant (Phase 5) and the$fieldsarray (email, phone, address) in the Checkout Validation (Phase 8) inreferences/detailed-guide.md. \n - Boundary markers: The code snippets lack clear delimiters (e.g., XML tags, triple backticks) or specific instructions for the model to ignore potential commands embedded within the user-supplied text. \n
- Capability inventory: The AI Shopping Assistant can be manipulated to output arbitrary text, and the Checkout Validation can be bypassed to influence order success/failure. \n
- Sanitization: While the code uses standard WordPress sanitization functions like
sanitize_text_field()andsanitize_email(), these are designed to prevent XSS and SQL injection, not semantic prompt injection. \n- [DATA_EXFILTRATION]: The provided code snippets for fraud detection (Phase 3) and checkout validation (Phase 8) collect sensitive PII (customer email, shipping address, phone number) and transmit it to an external AI service via thewp_ai_client_promptfunction. This represents a data exposure surface, though it is consistent with the skill's stated purpose of implementing AI connectors.
Audit Metadata