The Agent Skills Directory

Verified Accuracy Protocol: The skill implements a mandatory adversarial review step where a fresh subagent independently verifies every new claim against source documentation. This 'skeptical reviewer' persona is specifically instructed to flag unsourced or misleading information, which is a best practice for maintaining content integrity.
Controlled Execution Environment: Although the skill analyzes and transforms user-provided documentation, it does not perform any administrative actions, modify system configurations, or write files to the disk. The output is presented purely as a comparison within the conversation.
Safe Documentation Examples: The provided reference materials contain code snippets and API examples (e.g., curl commands) that use standard placeholders like {zone_id} and {token}. These are correctly presented as instructional templates for the user rather than executable instructions for the agent.
Clear Capability Boundaries: The skill explicitly defines its scope, targeting overview and conceptual pages while advising against use on sensitive configuration or developer-focused API references where technical jargon is the primary content. This reduced scope minimizes the risk of accidental misinterpretation of complex technical parameters.
Authoritative Source Verification: The instructions direct the agent to cross-reference any new information specifically against the repository's documentation, rather than relying on internal knowledge or external websites, which mitigates the risk of supply-chain or external data poisoning.

eli5