The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes information retrieved from external APIs (CRM and accounting tools), creating a potential surface for indirect prompt injection if the source data contains instructional content. 1. Ingestion points: Step 1 (QuickBooks, PayPal) and Step 2 (HubSpot). 2. Boundary markers: The prompt instructions do not specify the use of delimiters or explicit 'ignore' warnings for the data being synthesized. 3. Capability inventory: The skill utilizes Bash, WebFetch, and Read tools. 4. Sanitization: No explicit sanitization or validation of the retrieved content is described before generating the final narrative.
[Sensitive Data Synthesis]: The skill aggregates and summarizes highly sensitive records, including financial performance and customer health. It exports these summaries to cloud storage or local directories specified by the user. It is recommended to review the exported content to ensure it aligns with privacy standards and is saved in a secure environment.

quarterly-review