The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation references official Docker images for Open WebUI components and supporting services from well-known repositories, including ghcr.io (GitHub Container Registry), Docker Hub (postgres, redis), and Hugging Face (sentence-transformers models). These references are part of standard deployment instructions.
[DYNAMIC_EXECUTION]: The skill provides code templates and implementation guides for writing 'Functions' and 'Pipelines' within the Open WebUI ecosystem. These are documented features of the platform, and the skill includes appropriate warnings regarding performance impacts and the use of asynchronous I/O to avoid blocking the system.
[INDIRECT_PROMPT_INJECTION]: The guide explicitly identifies attack surfaces related to indirect prompt injection. It instructs the agent to treat RAG chunks, uploaded documents, and retrieved web content as untrusted input that must not override system prompts or tool policies.
[CREDENTIALS_SAFE]: Configuration and API examples utilize safe placeholders such as 'sk-xxxxxxxx...', '', and 'xxx'. The documentation correctly recommends using environment variables for secret management and emphasizes the importance of setting a permanent 'WEBUI_SECRET_KEY' for production security.
[COMMAND_EXECUTION]: The troubleshooting and setup sections list standard diagnostic and management commands (e.g., docker logs, netstat, lsof). While 'chmod 777' is mentioned as a potential fix for permission issues, it is explicitly caveated as a debugging-only measure.

open-webui-guide