Compress AI memory by classifying entries and generating summary files.
npx skills add https://github.com/bketelsen/truenas-mcp --skill memory-compact
Please help me install this Agent Skill. Skill: memory-compact Source: https://github.com/bketelsen/truenas-mcp/tree/main/.gaai/core/skills/cross/memory-compact Command: npx skills add https://github.com/bketelsen/truenas-mcp --skill memory-compact
This Skill addresses high context window pressure and overgrown memory by performing a targeted, single-pass compression of AI memory.
Run the memory-compact skill to compress the AI's memory.
High-intent search queries and answers about installing and using this skill.
To compress AI agent memory, you can run a single-pass compression process that classifies memory entries by durability, generates summary files for compacted categories, and archives original ephemeral or superseded durable entries to reduce context window pressure.
You need to archive ephemeral memory during standard compression, while durable memory entries are only archived if explicitly superseded. A classification gate ensures durable memory is protected and only superseded entries are removed.
The best way to summarize AI memory during a complex task is to use a single-pass compression that generates concise summary files for compacted categories, preserving key decisions and constraints without losing critical information.
Yes, you can trigger emergency memory compression mid-task when token limits are being approached. This single-pass compression quickly reduces the AI memory footprint by archiving ephemeral entries and generating category summaries.
After compacting AI agent memory, the memory index is updated automatically upon completion of the compression process. This ensures the index reflects the newly generated summary files and the archived original entries.
Yes, memory compression works without losing critical constraints by generating concise summary files for compacted categories. This preserves key decisions and constraints while archiving original ephemeral or explicitly superseded durable entries.
Fr-e-dCompress AI agent memory by category and archive originals.
BrennonTWilliamsCompact and summarize session memory to reduce context footprint.
mapachekurtMonitor Claude conversation token usage and issue warnings at critical thresholds.
phuonghxOptimize AI agent context windows by selectively loading tokens within a budget.
