The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Both 'scripts/agent.py' and 'scripts/process.py' include code that explicitly bypasses SSL/TLS certificate verification using the 'curl -k' flag and 'ssl.CERT_NONE' respectively. This configuration allows connections to be intercepted by a Man-in-the-Middle (MITM) attack, potentially compromising the PAN-OS API key.
[CREDENTIALS_UNSAFE]: Sensitive API keys are transmitted as plaintext query parameters in URL GET requests. This exposure increases the risk of credential leakage via network logging infrastructure and proxy logs.
[COMMAND_EXECUTION]: The 'scripts/agent.py' utility uses 'subprocess.run' to execute the 'curl' command. While implemented without a shell wrapper, this establishes a dependency on the local system's 'curl' binary and environment.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect injection by ingesting XML data from network firewalls and processing it into reports. 1. Ingestion points: 'scripts/agent.py' and 'scripts/process.py' fetch XML from the firewall API. 2. Boundary markers: No delimiters or instructions are used to distinguish external data from agent instructions. 3. Capability inventory: The skill can execute subprocesses and write JSON reports to the local filesystem. 4. Sanitization: The scripts rely on basic XML parsing but lack explicit validation or sanitization of the data values before they are used in reporting.

implementing-next-generation-firewall-with-palo-alto