The Agent Skills Directory

[SAFE]: The skill provides documentation and examples for using the ae-cli tool to manage project configurations, tasks, and campaign activities. The commands and resources described (e.g., ae-cli, engage-scene, engage-flow) are specific to the Agentic Engine platform infrastructure owned by the vendor 'thinkingaiagenticengine'. No malicious patterns, hardcoded secrets, or unauthorized network operations were detected.
[INDIRECT_PROMPT_INJECTION]: The skill manages a large volume of untrusted data from the AE platform, creating a potential surface for indirect prompt injection.
Ingestion points: The agent retrieves complex JSON objects and reports via commands like engage-task task get, engage-flow flow get, and engage-activity activity info-list (described in SKILL.md and various references).
Boundary markers: The skill explicitly instructs the agent to 'Never fabricate or guess resource names' and to 'Read the matching command reference and discover real project metadata first', which serves as a procedural boundary against malicious input influencing agent behavior.
Capability inventory: The agent has extensive write capabilities across the platform, including saving flows and tasks (engage-flow flow save, engage-task task save), managing lifecycle states (engage-task task manage), and configuring campaign activities.
Sanitization: The skill mandates the use of semantic definition contracts for audience and trigger logic, which are compiled and validated by the backend service rather than being directly interpreted by the agent, reducing the risk of malicious payload execution.

ae-engage