skills/kdocs-app/kdocs-skill/kdocs/Gen Agent Trust Hub

kdocs

Warn

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPERSISTENCEPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads binary components and checksums from wpsai.wpscdn.cn, which is the official content delivery network for WPS.
  • [REMOTE_CODE_EXECUTION]: The setup scripts download and run a platform-specific binary executable (kdocs-cli). This is an essential component for document processing but involves executing external code.
  • [PERSISTENCE]: The setup.ps1 script for Windows automatically modifies the host's User PATH environment variable. The shell setup scripts provide explicit instructions for users to modify their bash or zsh configuration files to maintain tool access across sessions.
  • [PRIVILEGE_ESCALATION]: Installation instructions recommend using the -ExecutionPolicy Bypass flag in PowerShell to permit script execution, which bypasses default security restrictions.
  • [COMMAND_EXECUTION]: The agent uses the kdocs-cli tool to perform document management tasks. This allows the agent to execute a wide variety of cloud-based file operations based on user input.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it ingests untrusted data from cloud documents via read_file (references/drive/read_and_download.md). While boundary markers for deletion are present in SKILL.md, they are absent for general document content interpretation. The skill provides extensive capabilities via kdocs-cli, and no explicit sanitization of external document content is defined.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 22, 2026, 07:29 AM