HarnessClaw Engine: A Go-based LLM Programming Assistant with WebSocket and Tool Calling

HarnessClaw Engine is an LLM programming assistant engine developed in Go. It leverages the WebSocket protocol (v1.4) to facilitate multi-turn dialogues and supports a comprehensive tool-calling mechanism. Key features include a 5-phase query loop for preprocessing, LLM streaming calls, error recovery, parallel/serial tool execution, and continuation checks. The engine also incorporates a 6-step permission pipeline with various modes, a skill system that loads from `SKILL.md` files, and multi-provider support for LLMs like Anthropic, OpenAI, Bedrock, and Vertex. Additionally, it includes context compaction for managing token usage and thread-safe session management. The project recently released its v0.0.18 update.