The Agent Skills Directory

Indirect Prompt Injection Surface: The skill ingests and processes data from external sources including QuickBooks, PayPal, and HubSpot. This creates a potential surface where untrusted data from these platforms could influence the agent's behavior if it contains instructional patterns.
Ingestion points: Data is pulled in Step 1 (QuickBooks, PayPal) and Step 2 (HubSpot).
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the fetched data.
Capability inventory: The skill utilizes Bash for file generation and WebFetch for data retrieval.
Sanitization: There are no specific instructions for sanitizing or escaping the content retrieved from external APIs before it is used in the narrative generation.
Handling of Sensitive Information: The skill is specifically designed to access and synthesize highly sensitive business data, including P&L statements, revenue validation from payment processors, and customer health metrics from CRM systems. It exports this data to external storage providers (Google Drive/OneDrive) or the local desktop, which warrants review to ensure data is handled according to organizational policy.
Shell Command Usage: The skill is granted access to the Bash tool, which is used in Step 6 to generate and export PDF or presentation files. Executing shell commands to handle sensitive data exports is a standard practice for this functionality but requires monitoring for potential command injection if parameters were influenced by external inputs.

quarterly-review