Compress OpenClaw conversations with Cornell-MapReduce while preserving IDs, dates, and amounts.
npx skills add https://github.com/wjlmrzd/OpenClaw-Wren --skill deflate
Please help me install this Agent Skill. Skill: deflate Source: https://github.com/wjlmrzd/OpenClaw-Wren/tree/main/skills/deflate Command: npx skills add https://github.com/wjlmrzd/OpenClaw-Wren --skill deflate
OpenClaw conversations often grow without bound, driving token costs and slowing responses. Deflate provides a structured, lossless compression approach that preserves critical data while reducing overall context length.
Run the Deflate skill to compress your current conversation while preserving critical data.
High-intent search queries and answers about installing and using this skill.
To compress OpenClaw conversation context and reduce token cost, the Deflate skill applies the Cornell-MapReduce method to compress topics independently while preserving critical identifiers, dates, and amounts for auditability.
Yes, context compression preserves IDs, dates, and amounts for auditability. The lossless data preservation approach ensures critical identifiers are retained when compressing each topic independently during a session.
You should use /new instead of /compact when memory holds essential data that cannot be lost. The skill recommends /new to save tokens and cost while ensuring safe session transitions via MEMORY.md flush workflows.
The best way to manage multi-topic sessions without losing key data is tracking topics across the session, marking completed topics, and compressing them independently to preserve critical identifiers while reducing overall context length.
Yes, you can report compression quality after compressing a conversation. The skill enforces lossless data preservation and reports compression quality to verify that critical data is retained while reducing token cost.
Fandry96Compress agent conversation history into structured summaries with artifact tracking.
AreteDriverOrganize Claude Code session memory files and handoff documents.
AngelMaldonadoStore and organize project memories in a SQLite-backed memory store.
ken-cavanagh-gleanCompress long-running agent session context with anchored iterative summarization.
