The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard system utilities such as dig, whois, curl, and openssl for reconnaissance and verification of infrastructure state. These commands are used according to best practices and are necessary for the skill's intended purpose of domain and SSL management.
[DATA_EXPOSURE]: While the skill requires sensitive environment variables (CLOUDFLARE_API_TOKEN, CLOUDFLARE_ACCOUNT_ID), it includes specific safety rails that prohibit the agent from echoing or logging these credentials, reducing the risk of accidental exposure.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided domains and incorporates them into command-line operations, creating a theoretical attack surface for prompt injection via DNS records or HTTP headers. However, the skill mitigates this risk by requiring a structured workflow of planning, dry-runs, and explicit human-in-the-loop confirmation before any execution occurs.
Ingestion points: User-provided domain names in SKILL.md workflow.
Boundary markers: None explicitly defined for untrusted data isolation.
Capability inventory: Uses dig, whois, curl, openssl, and cf api commands across the workflow steps defined in SKILL.md.
Sanitization: Relies on mandatory user confirmation and dry-run outputs rather than programmatic string escaping.

infra-automation