The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by ingesting untrusted data from the pfSense API and processing it for report generation.
Ingestion points: scripts/agent.py fetches firewall rules via get_firewall_rules() and system logs via get_firewall_logs(). These sources contain strings, such as rule descriptions and log messages, that can be influenced by network traffic or administrative actions.
Boundary markers: The generated report (pfsense_report.json) does not employ delimiters or explicit instructions to inform downstream agents that the ingested content should be treated as untrusted data.
Capability inventory: The agent.py script has broad write capabilities (post, put, delete) to the pfSense API, enabling it to create, modify, or delete firewall rules and NAT configurations.
Sanitization: No sanitization, escaping, or validation of the rule descriptions or log entries is performed before they are included in the JSON output report.
[EXTERNAL_DOWNLOADS]: The skill references a remote IP blocklist from a well-known community provider to manage firewall aliases.
The SKILL.md file defines a BlockedCountries alias that fetches data from https://www.ipdeny.com/ipblocks/data/aggregated/cn-aggregated.zone. This is an established service for network security professionals and is considered a safe resource for its intended purpose.

configuring-pfsense-firewall-rules