The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it integrates untrusted data from the Zscaler environment with high-privilege capabilities.
Ingestion points: The skill fetches external data using tools such as list_firewall_rules, list_url_filtering_rules, list_dlp_dictionaries, and list_vpn_credentials.
Boundary markers: There are no boundary markers or instructions defined to prevent the agent from interpreting content within the API responses as commands.
Capability inventory: The skill includes high-impact tools like create_firewall_rule, update_firewall_rule, and delete_firewall_rule which can modify the network's security configuration.
Sanitization: No evidence of data sanitization or validation is present in the skill instructions.

zscaler-zia