Audit and clean LanceDB vector memory for Clawdbot deployments.
npx skills add https://github.com/xintuchain/tongtong --skill memory-hygiene-xintuchain
Please help me install this Agent Skill. Skill: memory-hygiene Source: https://github.com/xintuchain/tongtong/tree/main/skills/memory-hygiene Command: npx skills add https://github.com/xintuchain/tongtong --skill memory-hygiene-xintuchain
Prevents bloated LanceDB vector memory and excessive token usage by auditing, cleaning, and reseeding stored memories so the agent only recalls relevant facts.
Run an audit to list top memory entries, disable autoCapture, wipe the LanceDB store, and reseed key facts from MEMORY.md to restore a concise memory.
High-intent search queries and answers about installing and using this skill.
To clean a cluttered LanceDB vector memory, audit stored entries to identify duplicates and low-value data, safely wipe the LanceDB store, and reseed it with curated key facts to reduce token waste.
Clawdbot agents recall irrelevant logs when autoCapture saves noisy, low-value entries during long-running operations, which you can prevent by disabling autoCapture and reseeding only important facts.
Yes, you can automate LanceDB memory maintenance by disabling autoCapture and scheduling monthly maintenance jobs to wipe and reseed memory automatically.
Managing Clawdbot memory requires filesystem permission to manage the LanceDB store directory, access to the gateway configuration, and the ability to run memory_recall, memory_store, and gateway restarts.
After wiping the LanceDB store, reseed memory by storing curated key facts such as user preferences, important decisions, and lessons learned from your MEMORY.md file to restore efficient behavior.
Qiue-GConfigure memory-lancedb-pro MCP plugin for OpenClaw with lifecycle management.
ProyectoG007Audit, wipe, and reseed Clawdbot's LanceDB vector memory.
gaos6ePersist AI agent context across sessions using LanceDB vectors and Git-Notes.
kimasplundRank vector-store memories by learned q-values to improve retrieval relevance.
