dingtalk-knowledge-search

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes standard utilities like pdftotext and the trusted agent-browser tool to manage document conversion and browser sessions. Subprocess calls are handled safely without shell interpretation and use validated paths.
  • [EXTERNAL_DOWNLOADS]: Communication is limited to official DingTalk API endpoints. Dependencies are sourced from trusted organizations like Vercel Labs.
  • [CREDENTIALS_UNSAFE]: Sensitive API keys and tokens are stored in a user-local configuration file (~/.dingtalk-skills/config), following industry-standard practices for local secret management.
  • [PROMPT_INJECTION]: Evaluated the risk of indirect prompt injection from processed DingTalk content.
  • Ingestion points: External document data retrieved via DingTalk APIs and browser extraction in cli.py.
  • Boundary markers: Usage of Markdown comments to separate document metadata from content.
  • Capability inventory: Execution of specific conversion tools and browser automation, with localized file-writing capabilities.
  • Sanitization: Employs path traversal protection in the resource proxy and magic byte validation for downloaded files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 02:14 PM
Security Audit — agent-trust-hub — dingtalk-knowledge-search