The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx to fetch and execute instructions and tools from the vendor's npm package (@nomadamas/k-skill) at runtime, as specified in the SKILL.md file.
[EXTERNAL_DOWNLOADS]: The skill downloads the katok CLI from the vendor's GitHub repository using Homebrew or Cargo during the setup process, as described in instruction.md.
[COMMAND_EXECUTION]: The agent is instructed to run several shell commands including npx, brew, cargo, and the katok CLI to manage data synchronization, search message archives, and request macOS system permissions (Full Disk Access).
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted chat data from local KakaoTalk archives, creating a vulnerability surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through the katok sync, katok search, and katok chunk get commands that read local KakaoTalk databases as referenced in the workflow section of instruction.md.
Boundary markers: There are no defined boundary markers or explicit instructions provided to the agent to ignore or delimit embedded commands within the retrieved chat content.
Capability inventory: The skill executes local database synchronization and various search operations via the katok CLI.
Sanitization: The instructions do not specify any sanitization, filtering, or escaping for the retrieved message content before it is processed by the agent.

kakaotalk-mac