The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes DNS record values (such as TXT or CNAME records) fetched from the DreamHost API. Because these records can be set by third parties, they provide a potential vector for indirect prompt injection.
Ingestion points: External DNS data enters the context in dreamhost_skill.py through the dns-list_records API command.
Boundary markers: The SKILL.md file contains a 'CRITICAL: DNS Modification Confirmation Required' section that explicitly instructs the agent to obtain human approval before applying changes, which acts as a safety guardrail.
Capability inventory: The skill has the ability to modify DNS records via the DreamHost API and perform lookups using the system's dig tool.
Sanitization: The Python script handles record data as strings and does not perform specific sanitization for malicious instruction patterns inside the record content.
[COMMAND_EXECUTION]: The cmd_check_dns function in dreamhost_skill.py executes the dig utility using subprocess.run. The implementation correctly uses an argument list instead of a shell string, preventing shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to api.dreamhost.com. This is the official endpoint for a well-known service (DreamHost) and is used for its intended purpose of DNS management.

dreamhost-skill