openkakao-cli

Warn

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSPERSISTENCE
Full Analysis
  • [DATA_EXFILTRATION]: The skill manages highly sensitive data, including authentication tokens extracted via the login --save command and private chat content from the local database. It provides built-in functionality to send this data to external, user-defined services through the --webhook-url flag.\n- [COMMAND_EXECUTION]: The watch --hook-cmd parameter allows for the execution of arbitrary shell commands in response to new message events. This functionality creates a risk of command injection if incoming message content is interpolated into the command string without proper sanitization.\n- [INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection by ingesting untrusted real-time KakaoTalk messages.\n
  • Ingestion points: Incoming chat messages processed via the watch command in SKILL.md.\n
  • Boundary markers: None identified; message content is processed directly by hooks and webhooks.\n
  • Capability inventory: The tool can send, delete, and edit messages, as well as execute arbitrary local commands via shell hooks.\n
  • Sanitization: There is no evidence of input validation or sanitization for message data before it is used in automated hooks or webhooks.\n- [EXTERNAL_DOWNLOADS]: The documentation instructs the user to install the CLI tool from a third-party GitHub repository and Homebrew tap (JungHoonGhae/openkakao). While these resources are provided by the skill's author, they involve executing external code on the host system.\n- [PERSISTENCE]: The skill documentation suggests patterns for maintaining persistence by integrating the CLI tool with system services such as launchd and cron for unattended background operations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 31, 2026, 06:30 PM