The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of untrusted data which could be used to influence agent behavior through configuration poisoning.
Ingestion points: SKILL.md defines commands such as hcloud firewall create --rules-file <file> and hcloud firewall replace-rules --rules-file <file> that read from external JSON files or stdin.
Boundary markers: Absent. The documentation does not provide instructions to the agent to treat these external files as untrusted or to validate their contents.
Capability inventory: The skill has the capability to perform wide-ranging network configuration changes (firewalls, load balancers, network routes) using the hcloud CLI.
Sanitization: Absent. There is no evidence of input validation or escaping for the data read from external rule files.
[COMMAND_EXECUTION]: The skill provides numerous templates for executing shell commands via the hcloud CLI tool.
Evidence: The commands in SKILL.md incorporate many user-supplied placeholders like <name>, <cidr>, and <ips>. This creates a surface for command injection if the agent does not properly sanitize these values before execution.

hcloud-networking