The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes Azure AD audit logs and sign-in data via the Microsoft Graph API. As this data originates from external systems (Azure AD), it represents a surface for indirect prompt injection if an attacker can manipulate log entries (such as application names, user principal names, or IP addresses) to influence the agent's behavior. * Ingestion points: query_audit_logs and query_signin_logs functions in scripts/agent.py. * Boundary markers: None explicitly used in the log processing logic to delimit external data. * Capability inventory: The script performs network reads via requests.get and writes the resulting report to a local JSON file. * Sanitization: The script focuses on extracting specific JSON fields and generating a structured report without explicit sanitization of the values.
[EXTERNAL_DOWNLOADS]: The scripts/agent.py script communicates with login.microsoftonline.com and graph.microsoft.com. These network operations target official Microsoft services and are necessary for the skill's functionality to retrieve Azure security logs.

detecting-azure-lateral-movement