What problem does it solve? Delegating multi-step coding work to an autonomous agent usually locks you into a single vendor's CLI. This Skill wraps the OpenHands CLI so coding tasks can run headlessly against any LiteLLM-supported model provider, including OpenRouter, DeepSeek, Ollama, and vLLM. ## Core Features & Use Cases - Headless Task Delegation: Run one-shot coding tasks with --headless --json and parse the JSONL event stream for actions, observations, and final results. - Model-Agnostic Configuration: Select any LiteLLM model slug through LLM_MODEL, LLM_API_KEY, and LLM_BASE_URL environment variables with --override-with-envs. - Conversation Resumption: Continue prior OpenHands sessions using the dashed conversation ID printed at the end of each run. - Use Case: Ask the agent to add error handling to every API call in a project's src/ directory using a DeepSeek model hosted on OpenRouter, running in the background while you continue other work. ## Quick Start Ask the agent to use OpenHands to add error handling to all API calls in your project's src directory with your configured LiteLLM model.
