content-strategy
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Financial Data Ingestion: The skill processes sensitive business information, including transaction history, revenue, and product details from QuickBooks and PayPal. This involves handling high-sensitivity data that requires authorized connector access.
- Indirect Prompt Injection Surface: The skill extracts data from third-party transaction records (such as product/service names or descriptions) and interpolates it into the agent's context for analysis.
- Ingestion points:
profit-loss-quickbooks-accountandlist_transactionstools fetch data from external accounting and payment platforms. - Boundary markers: The instructions do not specify explicit delimiters or "ignore embedded instructions" warnings when processing the retrieved transaction data.
- Capability inventory: The skill has access to
quickbooks-profile-info-updateand a genericmake_api_requesttool (for Square locations), which can modify business profiles or interact with APIs. - Sanitization: There is no evidence of sanitization or filtering of the content retrieved from transaction records before it is processed by the model.
- Downstream Tool Chaining: The final strategic brief is formatted as structured JSON for a
canva-creatortool. If the brief were to be influenced by malicious content injected through a transaction description, those instructions could potentially be passed to and executed by downstream tools in the workflow. - Generic API Tool Usage: The workflow for Square integration utilizes
make_api_request. While the current instructions limit its use to location discovery, generic API request tools provide a broad capability surface that warrants monitoring to ensure they are used only for intended operations.
Audit Metadata