The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data retrieved from the Nessus scanner, which could be manipulated by an attacker to influence the agent's behavior.
Ingestion points: The get_scan_details and check_auth_status methods in scripts/agent.py fetch vulnerability names, hostnames, and plugin outputs from the Nessus REST API.
Boundary markers: SKILL.md does not define delimiters or provide instructions to the agent to treat data from the scanner as potentially untrusted.
Capability inventory: The script's primary functions are data retrieval and console reporting; it does not currently possess high-risk capabilities like writing to the filesystem or executing arbitrary shell commands based on scan results.
Sanitization: The script does not perform sanitization or validation of the hostnames or vulnerability descriptions retrieved from the scanner.
[DATA_EXFILTRATION]: The script scripts/agent.py disables TLS/SSL certificate verification by setting self.session.verify = False and suppressing InsecureRequestWarning. While common in internal environments using self-signed certificates, this practice allows an attacker positioned on the network to intercept Nessus API keys and sensitive scan results via a Man-in-the-Middle (MitM) attack.

performing-vulnerability-scanning-with-nessus