The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/agent.py utilizes the pickle.load() function to deserialize machine learning models from a path provided via the --dga-model command-line argument. Because the pickle format is capable of executing arbitrary code during the deserialization process, loading an untrusted file through this method poses a significant security risk.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted DNS log data which may contain attacker-controlled strings in subdomains or TXT records. This creates a vulnerability surface where malicious content within the logs could attempt to influence the agent's behavior or decision-making process when reviewing analysis results.
Ingestion points: Data is ingested from external files in scripts/agent.py through the parse_zeek_dns_log, parse_suricata_eve, and parse_csv_dns functions.
Boundary markers: There are no explicit instructions or delimiters used when parsing the logs to ensure the agent treats the content strictly as data rather than instructions.
Capability inventory: The script possesses the capability to read from and write to the local file system and uses standard data processing libraries like numpy and scikit-learn.
Sanitization: The ingestion logic does not perform rigorous sanitization or filtering of the queried domains or record responses beyond basic character encoding handling.

detecting-command-and-control-over-dns