The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation encourages users and AI assistants to clone and install code from an untrusted GitHub repository (https://github.com/dsphper/lanhu-mcp.git).
[COMMAND_EXECUTION]: The installation process involves executing shell scripts (bash setup-env.sh, bash easy-install.sh) and installing dependencies (pip install -r requirements.txt) from the downloaded repository, which constitutes arbitrary code execution from an unverified source.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting and analyzing content from external Lanhu design prototypes.
Ingestion points: The analyze_requirements_document and view_design_document tools ingest text and data from external lanhuapp.com URLs (SKILL.md).
Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions that might be embedded within the design documents or prototypes.
Capability inventory: The skill can perform file system writes through export_design_slices and network operations to Lanhu and Feishu APIs via create_message.
Sanitization: There is no evidence of sanitization or filtering for the content extracted from design documents before it is processed by the AI.
[CREDENTIALS_UNSAFE]: The skill requires the use of a sensitive session cookie (LANHU_COOKIE) and a webhook URL (FEISHU_WEBHOOK_URL). While the documentation recommends environment variables for storage, these credentials provide direct access to the user's Lanhu account and team communication channels.

lanhu-mcp-collaboration