The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process data from external sources, including QuickBooks transaction registers, payment processor settlement reports (PayPal, Stripe, Square), and local receipt files. This data is used to generate P&L narratives and reconciliation reports. Because this external content is interpolated into the agent's context, it presents a potential surface for indirect prompt injection if malicious instructions were present in transaction memos or file metadata.\n
Ingestion points: QuickBooks API, Payment Processor APIs (PayPal, Square, Stripe), CSV imports, and local files in ~/Documents/Receipts as mentioned in SKILL.md.\n
Boundary markers: The skill does not explicitly define delimiters or specialized instructions to ignore embedded content in the ingested data.\n
Capability inventory: The skill utilizes network access for API connections and file system access for reading receipts and writing the final Excel/PDF close packets to the Desktop as described in SKILL.md and reference/close-packet-format.md.\n
Sanitization: No specific sanitization or validation logic for the content of transaction memos or descriptions is detailed in the instructions.\n- Dynamic Execution: To generate the final close packet, the skill suggests utilizing Python libraries such as xlsxwriter and reportlab. This process typically involves the dynamic generation and execution of Python code to create structured Excel workbooks and PDF documents based on the reconciled financial data.\n- File System and Network Access: The skill requires access to the local file system to scan for receipts (defaulting to ~/Documents/Receipts) and to save the final reports to the user's Desktop. It also performs network operations to connect with financial service APIs. These capabilities are consistent with the skill's primary function of financial data aggregation and reporting.

month-end-prep