浏览器与自动化
common-context-optimization
Maximize context window efficiency, reduce latency, and prevent lost-in-middle issues through strategic masking and compaction. Use when token budgets are tight, tool outputs overflow the context, conversations drift from intent, or latency spikes from cache misses.
HoangNguyen0403 · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add HoangNguyen0403/agent-skills-standard --skill common-context-optimization技能说明
name: common-context-optimization description: Maximize context window efficiency, reduce latency, and prevent lost-in-middle issues through strategic masking and compaction. Use when token budgets are tight, tool outputs overflow the context, conversations drift from intent, or latency spikes from cache misses. metadata: triggers: files: - '*.log' - 'chat-history.json' keywords: - reduce tokens - optimize context - summarize history - clear output
Priority: P1 (HIGH)
1. Observation Masking (Noise Reduction)
Problem: Large tool outputs (logs, JSON lists) overwhelm context and degrade reasoning. Solution: Replace raw output with semantic summaries after consumption.
- Identify outputs exceeding 50 lines or 1 KB.
- Extract critical data points immediately.
- Mask by rewriting history to replace raw data with summary placeholder.
- See
references/masking.mdfor patterns.
See implementation examples for masking patterns.
2. Context Compaction (State Preservation)
Problem: Long conversations drift from original intent. Solution: Recursive summarization that preserves State over Dialogue.
- Trigger compaction every 10 turns or 8k tokens.
- Compact:
- Keep: User Goal, Active Task, Current Errors, Key Decisions.
- Drop: Chat chit-chat, intermediate tool calls, corrected assumptions.
- Format: Update System Prompt or Memory File with compacted state.
- See
references/compaction.mdfor algorithms.
See implementation examples for compacted state format.
3. KV-Cache Awareness (Latency)
Goal: Maximize pre-fill cache hits.
- Static Prefix: Enforce strict ordering — System -> Tools -> RAG -> User.
- Append-Only: Never insert into middle of history; append new turns only.
References
- Observation Masking Patterns
- Compaction Algorithms
Anti-Patterns
- No raw tool dumps: Mask large outputs immediately after extracting data.
- No unbounded growth: Compact every 10 turns to preserve intent over dialogue.
- No middle insertions: Append-only history maximizes KV cache hits.