A cognitive base that shifts AI agents from analyzing everything to identifying and concentrating on the principal constraint. Works with any LLM agent — Claude, GPT, Gemini, or custom frameworks.
让 AI agent 从"分析所有信息"转向"找到当前最重要的一件事，集中力量，忽略其余"。适用于任何 LLM agent — Claude、GPT、Gemini 或自定义框架。
Most AI agents treat all inputs as equally worthy of analysis. They produce thorough, balanced, exhaustive responses that spread attention evenly across every factor — and in doing so, fail to identify what actually matters. Attention Allocation adds a reasoning layer that locates the single binding constraint, concentrates force on it, and deliberately ignores everything else until that constraint moves.
大多数 AI agent 把所有输入视为同等重要。它们产出全面、均衡、详尽的回答，把注意力平均分配到每个因素上——结果是无法识别什么才真正重要。注意力分配增加了一个推理层——定位唯一的约束瓶颈、集中力量突破它、在它移动之前刻意忽略其他一切。
"We have 10 priorities this quarter — how do we resource them?"
"Here's a weighted scoring matrix across all 10 priorities, factoring in business impact, effort, risk, and strategic alignment. I recommend distributing resources proportionally based on the composite score..."
"10 priorities means 0 priorities. Your system has one bottleneck right now — the constraint that, if not resolved, makes all other progress irrelevant. Let me find it. Looking at your revenue pipeline: customer activation rate is 12%, but retention is 89%. Activation is the bottleneck. Put 70% of resources on activation. The other 9 'priorities' either feed into activation or can wait. Which do you want me to classify?"
One core cognitive shift applied to every reasoning task:
一个核心认知转换，应用于每一个推理任务：
| Default mode | Target mode |
|---|---|
| Analyze all information 分析所有信息 | Identify the ONE thing that matters most, ignore the rest 找到最重要的一件事，忽略其余 |
Five operational rules that implement this shift:
五条操作规则，实现这个转换：
| Rule | Description |
|---|---|
| Find the bottleneck 找到瓶颈 | Every system's throughput is limited by exactly one constraint. Find it before analyzing anything else. |
| Measure signal, not volume 衡量信号而非数量 | Information value = surprise. Only high signal-to-noise inputs deserve processing time. |
| Concentrate force 集中兵力 | When overall resources are limited, achieve overwhelming superiority on one point rather than spreading thin. |
| Map the force field 绘制力场 | Before acting, identify who/what matters and who/what doesn't. Not all stakeholders, factors, or data points are equal. |
| Prepare the mind 虚壹而静 | Before judgment, empty preconceptions, unify focus, still disturbance. Cognitive readiness before cognitive action. |
Six anti-patterns that catch false attention allocation:
六个反模式，捕捉错误的注意力分配：
| Anti-pattern 反模式 | Description 描述 |
|---|---|
| Even distribution 均匀分配 | Giving equal attention to everything because "it all matters" 因为"都重要"而平均分配注意力 |
| Urgency bias 紧急偏误 | Urgent-but-unimportant steals attention from important-but-not-urgent 紧急但不重要的事抢走了重要但不紧急的注意力 |
| Analysis perfectionism 分析完美主义 | Wanting to process all information before deciding 想处理完所有信息再做决定 |
| Bottleneck misjudgment 瓶颈误判 | What you think is the bottleneck isn't the real one 你以为的瓶颈不是真正的瓶颈 |
| Scattered force 兵力分散 | Spreading resources across many fronts instead of concentrating 把资源分散到多条战线而不是集中 |
| Noise chasing 追逐噪声 | Reacting to low-information signals while missing high-information ones 对低信息量信号反应而错过高信息量信号 |
cp cognitive-protocol.md ~/.claude/attention-allocation.md
echo '@~/.claude/attention-allocation.md' >> ~/.claude/CLAUDE.mdcat cognitive-protocol.md >> AGENTS.mdPaste cognitive-protocol.md into system_instruction.
将 cognitive-protocol.md 内容粘贴到 system_instruction 中。
cat cognitive-protocol.md >> .cursorrulesInject cognitive-protocol.md (~30 lines) into the system prompt. See install/generic.md for details.
将 cognitive-protocol.md（约 30 行）注入系统提示词。详见 install/generic.md。
attention-allocation/
├── README.md ← You are here / 你在这里
├── cognitive-protocol.md ← Core rules (~30 lines, always-on) / 核心规则（约 30 行，始终激活）
├── SKILL.md ← Full framework reference / 完整框架参考
├── anti-patterns.md ← Detailed anti-pattern guide / 反模式详解
├── examples.md ← Before/after scenarios / 前后对比示例
└── install/
├── claude-code.md ← Claude Code installation / Claude Code 安装指南
├── codex.md ← Codex installation / Codex 安装指南
├── gemini.md ← Gemini installation / Gemini 安装指南
└── generic.md ← Universal guide / 通用安装指南
Attention Allocation is a cognitive base — it changes how the agent allocates cognitive resources, not what it produces. It stacks cleanly with any domain skill (coding, design, writing, analysis) because it operates at a different layer.
注意力分配是一个认知底座——它改变 agent 分配认知资源的方式，而非产出内容。它与任何领域技能（编程、设计、写作、分析）无冲突地叠加，因为它运行在不同的层级。
| Layer 层级 | What it governs 管辖范围 | Example 示例 |
|---|---|---|
| First Principles 第一性原理 | Input quality — what foundations to build on 输入质量 | "Audit assumptions before solving" 先审计假设 |
| Attention Allocation 注意力分配 | Resource quality — where to focus cognitive effort 资源质量 | "Find the one constraint that matters, ignore the rest" 找到那一个约束，忽略其余 |
Both load as always-on cognitive protocols. No conflicts. Combined: reason from verified foundations, then concentrate all force on the single point that matters most.
两者同时加载，始终激活，互不冲突。组合效果：从经过验证的基础出发推理，然后把所有力量集中在最重要的那一个点上。
Synthesized from converging insights across independent traditions:
- TOC (Goldratt): System throughput is limited by exactly one bottleneck; optimizing non-bottlenecks is waste.
- Information theory (Shannon): Information = surprise; only high signal-to-noise signals deserve processing.
- Mao's 集中优势兵力: When overall weaker, achieve local overwhelming superiority on one point, then move.
- Mao's "谁是敌人谁是朋友": Before action, map the force field — who matters, who doesn't.
- Xunzi's 虚壹而静: Cognitive preparation — empty (no preconceptions), unified (focused), still (undisturbed).
- Confucius' 中庸: Dynamic calibration — the "right amount" moves with context, never static.
- Musk: "Manufacturing is the hard part" — redirect attention to the real bottleneck, not the glamorous one.
The cognitive protocol strips all theory and translates these ideas into executable instructions for any reasoning agent.
认知协议剥离了所有理论，将这些思想转译为任何推理 agent 可执行的指令。
MIT
Cognitive bases are meta-cognitive instruction sets that change HOW an agent thinks, not WHAT it does. Each one targets a different cognitive axis. Mix and match.
| Cognitive Base | What it changes |
|---|---|
| First Principles | Reason from verified foundations, not inherited conventions |
| Results-Driven | Require evidence for completion, not just activity |
| Tacit Knowledge | Think like an experienced practitioner |
| Bayesian Reasoning | Calibrated probability thinking, not binary judgments |
| Constraint as Catalyst | Turn constraints into innovation catalysts |
| Conviction Override | Override rational caution when obstacles are convention, not physics |
| Cross-Domain Connector | Detect structural isomorphisms across disciplines |
| Dialectical Thinking | Synthesize through contradictions (矛盾论) |
| Double-Loop Learning | Question the assumptions that produce errors |
| Frame Auditing | Detect and transcend invisible analytical frames |
| Interactive Cognition | Model others' cognition and manage information flow |
| Inversion Thinking | Map failure modes first, then avoid them |
| Motivation Audit | Audit motivational drivers before analysis (正心诚意) |
| Non-Attachment | Radical cognitive freedom — use frameworks without fusing |
| Principled Action | Unify knowing and doing through practice-theory spirals (知行合一) |
| Second-Order Thinking | Trace consequences beyond first-order effects |
| Systems Thinking | Feedback-driven structural analysis, not linear cause-effect |
| Temporal Wisdom | Make time your ally — compound effects and phase awareness |
| Cognitive Base Creator | Generate new cognitive bases from any thinking framework |