long-context-v2

Minimize startup context by loading MEMORY-LITE.md instead of full MEMORY.md. Use for session bootstrap optimization.

SKILL.md

GitHub

Long Context v2.0.0

创建者: 思捷娅科技 (SJYKJ)/zhaog100


MIT License

Copyright (c) 2026 思捷娅科技 (SJYKJ)

免费使用、修改和重新分发时,需注明出处。


Strategy: Startup Optimization

File Hierarchy

File Size Purpose Loaded When
MEMORY-LITE.md ~2KB Core identity, workflows, tracked PRs Every session startup
MEMORY.md ~10KB Full history, detailed logs Only when needed via memory_search
memory/YYYY-MM-DD.md Variable Daily raw notes Only when searching specific dates

How It Works

  1. Startup: Agent loads MEMORY-LITE.md (not full MEMORY.md)
  2. Deep recall: Use memory_search to find specific information
  3. Detailed read: Use memory_get to fetch specific lines from MEMORY.md

Token Savings

  • Before: MEMORY.md ~28KB → ~7000 tokens
  • After: MEMORY-LITE.md ~2KB → ~500 tokens
  • Savings: ~6500 tokens per session startup

When to Use

  • Session starts with large context files
  • Need to recall historical events without loading full memory
  • Want to reduce initial token consumption

References

  • Compaction patterns: references/compaction.md
  • Memory search tips: references/search-tips.md