The Agent Skills Directory

[SAFE]: The skill explicitly instructs the agent to identify and exclude private or authenticated routes (such as /admin, /account, /checkout) to prevent sensitive data exposure in the semantic index.
[SAFE]: Privacy is prioritized by forbidding the transmission of site content or query text to third parties, relying instead on a pinned local embedding model.
[COMMAND_EXECUTION]: The skill utilizes the @nicolasakf/agentize CLI tool for legitimate operational tasks including initialization (agentize init), validation (agentize check), and building the search index (agentize build). These tools are recognized as legitimate vendor resources from nicolasakf.
[EXTERNAL_DOWNLOADS]: The skill downloads a pinned local embedding model during its initial build process. This is a functional requirement for the semantic search feature and is documented as a local-only operation.
[SAFE]: No signs of obfuscation, malicious prompt injection, privilege escalation, or persistence mechanisms were detected. The verification steps include testing for the absence of excluded or internal API content, which is a security best practice.

agentize-websites