The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill describes functionality for 'Cloud Functions' and 'Scheduled Functions' that allow for the execution of user-supplied JavaScript. These scripts have access to network resources via axios/fetch and can perform lead mutations, representing a dynamic execution surface within the platform's environment.
Evidence: references/configuration-reference.md sections 6 and 7, and references/recipes.md Recipe 3 detail the use of custom JS for event-driven and cron-driven logic.
[INDIRECT_PROMPT_INJECTION]: The platform is designed to inject lead metadata directly into the agent's runtime prompt to provide context. If this metadata is sourced from untrusted external inputs like public forms or external CRM data, it could be used to inject instructions that influence the agent's behavior.
Ingestion points: SKILL.md specifies that lead metadata is supplied by 'external tools (a CRM, a form, the customer's own backend)'.
Boundary markers: The documentation does not specify the use of clear delimiters or 'ignore' instructions for the injected metadata blocks.
Capability inventory: The agent has extensive capabilities including lead status modification, network tool execution, and workflow handoffs across all reference scripts.
Sanitization: The documentation lacks mention of sanitization or validation of metadata content before it is interpolated into the prompt.

automation-architecture