The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to run npx -y @nomadamas/k-skill@0 to retrieve the primary instruction set and manage files. This executes remote code from an unverified source during the setup and usage phases of the skill.\n- [EXTERNAL_DOWNLOADS]: The instructions require downloading and installing the katok CLI from a third-party GitHub repository (github.com/NomaDamas/katok) using Homebrew or Cargo. This source is not recognized as a trusted organization or well-known service.\n- [PRIVILEGE_ESCALATION]: The skill requests "Full Disk Access" permissions on macOS. This is a significant security boundary that, once granted, allows the tool to access protected user data across the entire system.\n- [DATA_EXFILTRATION]: The skill is designed to sync and process sensitive private communication history from local KakaoTalk archives. The use of unverified external tools for this purpose presents a high risk for unauthorized data access or exfiltration.\n- [COMMAND_EXECUTION]: The skill makes extensive use of shell commands through the katok CLI to perform database operations, searches, and permission management.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted chat data which may contain adversarial instructions designed to influence the agent.\n
Ingestion points: Processes local KakaoTalk message archives via the katok CLI (instruction.md).\n
Boundary markers: There are no explicit markers or instruction delimiters used when processing or presenting retrieved message content.\n
Capability inventory: The agent can execute CLI commands, sync databases, and retrieve full message chunks (instruction.md).\n
Sanitization: The skill lacks mechanisms to sanitize or filter the content of the messages before they are processed by the agent.

kakaotalk-mac