memory-setup

memory-setup helps users configure persistent memory for the Moltbot/Clawdbot agent so it stops behaving like a goldfish and retains context across sessions. It addresses the common problem of an agent forgetting prior conversations, repeating questions, and losing continuity on projects, by enabling memory search and a structured on-disk memory layout.

The skill walks through enabling a memorySearch block in the agent config (~/.clawdbot/clawdbot.json or moltbot.json) with options for provider, sources, index mode, minimum relevance score, and max results; creating a workspace memory structure (a curated MEMORY.md plus a memory/ tree for daily logs, projects, groups, and system notes); and initializing MEMORY.md and daily log templates. It documents each config setting with recommended values, offers embedding-provider choices (Voyage, OpenAI, or a local no-API option), and provides AGENTS.md instructions telling the agent to run memory_search and memory_get before answering questions about prior work. It also includes troubleshooting for search not working, irrelevant results, and provider errors, plus a gateway restart step and a verification test.

Target users are Moltbot/Clawdbot operators who want durable, searchable long-term memory and better project continuity. The skill only edits local config and creates local markdown files, and it references embedding API keys (VOYAGE_API_KEY, OPENAI_API_KEY) as standard environment-variable setup with a local no-key alternative. There is no data exfiltration or destructive behavior, so it is a benign configuration helper.