The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts/agent.py script disables SSL/TLS certificate verification by default during API interactions.
Evidence: The QRadarClient class initializes its SSL context with ssl.CERT_NONE whenever the verify_ssl parameter is False (the default state).
This insecure configuration allows for potential Man-in-the-Middle (MitM) attacks, which could result in the exposure of sensitive security events and the API authentication token passed in the headers.
[DYNAMIC_EXECUTION]: The script scripts/agent.py allows for the execution of arbitrary AQL queries passed through command-line arguments.
Evidence: The script collects all arguments after the action keyword using " ".join(sys.argv[2:]) and passes them directly to the client.search_aql method.
While this is functional for a CLI tool, it allows the agent to execute complex queries that could be manipulated by malicious instructions.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process log data from a SIEM, which constitutes an untrusted data source.
Ingestion points: Event data and flows retrieved from the QRadar API via the search_aql method in scripts/agent.py.
Boundary markers: The skill does not implement delimiters or safety instructions to separate the agent's logic from the potentially malicious content within the logs.
Capability inventory: The agent has the ability to make network requests, query the SIEM, and modify SIEM state (e.g., closing offenses).
Sanitization: No sanitization or validation is applied to the data returned from the SIEM before it is analyzed or presented.

correlating-security-events-in-qradar