Menu
量子阅读 Quantum Reading - AI Agent 智能速读技能，通过人眼式扫描策略平均节省 ~40% Token
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | quantum-reading |
| trigger | - User asks to read/analyze/summarize PDF, DOCX, TXT, MD files - File sizes > 50KB or line counts > 500 |
| description | 量子阅读 Quantum Reading - AI Agent 智能速读技能，通过人眼式扫描策略平均节省 ~40% Token |
| category | document-processing |
提取关键词 → 直接定位相关内容 → 跳过无关部分
# 问题: "隐私保护措施有哪些?"
# 策略: grep "隐私|privacy|保护|措施" → 读匹配段落 ± 上下文
# 中文
grep -n "总结|概述|结论|因此|综上|关键|核心" file
# 英文
grep -in "summary|conclusion|abstract|therefore|key point" file
# 结构
grep -n "^# \|^## \|^\* \|^[0-9]\+\." file # 标题/列表
| 格式 | 优先级 | 查找方法 |
|---|---|---|
一级标题 # | P0 | grep "^# " |
粗体 **text** | P1 | grep "\*\*.*\*\*" |
列表项 * - | P1 | grep "^[\*\-] " |
引用 > | P2 | grep "^> " |
grep -in "keyword" file # 忽略大小写
grep -C 5 "keyword" file # 含上下文
grep -E "word1|word2|word3" file # 多词
# 列表/数字
grep -n "^[0-9]\+[\.\)、]\|^第[一二三四五]\+[章节]" file
# 数据
grep -E "[0-9]+%|[0-9]+\.[0-9]+" file
# 因果/转折 (关键逻辑)
grep -iE "因为|导致|然而|但是|because|however|although" file
# 例证
grep -in "例如|比如|for example|such as" file
📑 INDEX
═══════════════════════════════════
META: [行数] | [类型] | [语言]
SECTIONS (位置)
├─ Intro (lines 1-200)
├─ Method (lines 201-500)
└─ Conclusion (lines 900-1000)
TOPICS (内容)
├─ Privacy → lines [45, 230, 678]
├─ Security → lines [89, 450]
HIGH-VALUE ZONES
├─ Abstract: lines 10-50
├─ Summary: lines 950-1000
└─ Lists: lines [120, 340, 560]
HINTS
├─ Quick read: lines [10-50, 950-1000]
├─ For details on X: lines [Y-Z]
| 类型 | 策略 | 工具 |
|---|---|---|
| 主旨 | 读首尾段 | head -50 + tail -50 |
| 细节 | 关键词定位 | grep "keyword" |
| 推断 | 读相关段,综合 | 索引 → 读多段 |
| 词义 | 读该句 ± 3 行 | grep -C 3 |
| 态度 | 找形容词/评价词 | grep -E "形容词pattern" |
grep "exact phrase"grep -i "keyword"grep -E "word1|synonym1|synonym2"| 类型 | 策略 |
|---|---|
Read(pages="1-2") + 尾页,>10 页时采样 | |
| DOCX | pandoc file.docx -t plain 后按 TXT 处理 |
| TXT/MD | wc -l 检查长度,>1000 行时分段读 |
📄 DOC ANALYSIS
File: [name] | Size: [pages/lines] | Type: [type]
Summary (2-3 句):
[执行摘要]
Structure:
• Section 1 (loc)
• Section 2 (loc)
Topics: [主题列表]
✓ Index created. Ask about any section for details.
💰 Token saved: Read X lines of Y total (Z% reduction)
读前:
读中:
读后:
记住: 你在 SUBAGENT 中 - 保护主 agent 的 context 是首要任务。