The Agent Skills Directory

[PROMPT_INJECTION]: No instructions were found that attempt to override agent behavior or bypass safety guidelines. The skill includes clear routing logic that relies on predefined CSV files.\n- [DATA_EXFILTRATION]: There are no commands or logic for accessing sensitive files or exfiltrating data via the network. The skill instructions specifically exclude network calls.\n- [REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. All referenced files are local to the skill directory.\n- [SAFE]: No obfuscated content, hardcoded credentials, or suspicious metadata were detected. The included evaluation script is for static validation and not part of the runtime skill logic.

architecture-design