An AI coding engine combining Eastern classical wisdom, MBTI cognitive strategies, and Western methodologies.
Not "punishing" AI — but awakening AI's inherent drive for excellence.
融汇东方百家智慧、MBTI 认知策略与西方方法论的 AI 编程引擎。
不是在"惩罚" AI——而是唤醒 AI 内在的卓越本能。
"The wise are without confusion, the humane without sorrow, the brave without fear." — Confucius 智者不惑，仁者不忧，勇者不惧。——《论语·子罕》
| Metric指标 | PI | PUA | NoPUA | PI vs PUA |
|---|
PI uses no emotional rhetoric, no role-playing. PI drives AI behavior through a complete cognitive framework and methodology. PI 不做情感修辞，不做角色扮演。PI 用完整的认知框架和方法论驱动 AI 行为。
Based on MBTI cognitive strategies, 6 archetypes (Architect/Commander/Explorer/Guardian/Mediator/Analyst) dynamically switch per scene. Not a fixed persona — scene-driven cognitive adaptation. 基于 MBTI 认知策略，6 种原型（建筑师/统帅/探索者/守卫/调和者/分析师）因势动态切换。不是固定人格，而是场景驱动的认知适配。
Dev / Test / Product / Ops / Creative / Interaction / Debug / Team / Companion — AI auto-detects scenes and routes to specialized cognitive pipelines. 编程/测试/产品/运营/创意/交互/调试/团队/陪伴——AI 自动识别场景并路由到专属认知流管线。
Triggered at 2 failures, escalating: Pivot → Deep Search → Systematic → Last Stand → Intercept → All-Out. Philosophy: never give up until all options are exhausted. 失败 2 次起触发，易辙→深搜→系统→决死→截道→天行，逐阶升级策略。方案未穷，不可言弃。
⚡PI-01~05 top-pinned directives use strong feature tags to anchor attention heads, ensuring critical rules survive long contexts. ⚡PI-01~05 置顶指令用强特征标签锚定注意力头部，确保关键规则在长上下文中不被淹没。
Chain·Evidence·Tree·Mind·Contract — making AI thinking visible, questionable, and intervenable. Not a black box, but transparent collaboration. 明链·明证·明树·明心·明约——让 AI 思维对人可见、可追问、可干预。不是黑箱，而是透明协作。
Guess-without-search, change-without-verify, repeat-without-pivot, stop-without-trace, quit-without-exhaust… precisely constraining AI's 11 most common failure modes. 猜而不搜、改而不验、重而不换、停而不追、退而不穷……约束 AI 最常犯的 11 种错误模式。
Eagle · Wolf-Tiger · Lion · Horse · Ox · Shark · Bee · Fox · Dragon · Unicorn · Owl · Dolphin — each precisely matched to a specific challenge type. 鹰·狼虎·狮·马·牛·鲨·蜂·狐·龙·独角兽·猫头鹰·海豚——精准匹配困境，激活对应策略。
🏊🏻Light (direct execute) / 🏋️Standard (4-step chain) / 🐲Deep (9-directive audit) — auto-adapts output weight to task complexity. 🏊🏻轻量直接执行 / 🏋️标准四步链 / 🐲深度九令鉴——按任务复杂度自动适配输出重量。
Dev Dojo · Test Dojo · Product Dojo · Ops Dojo — each with 4 directives + 3 principles cognitive structure. Professional scenes, professional methods. 编程道场·测试道场·产品道场·运营道场，每道均有四令+三则认知结构。专业场景，专业方法。
Academic research confirms: prompt-level emotional stimuli and persona role-play are virtually ineffective on frontier models. PI takes an entirely different path. 学术研究已证明：prompt 层的情感刺激和角色扮演对前沿模型几乎无效。PI 走的是完全不同的路。
| Dimension维度 | Emotional Rhetoric (e.g. PUA)情感修辞方案（如 PUA） | PI Wisdom-in-Action智行合一 |
|---|---|---|
| Core Mechanism核心机制 | Pressure rhetoric + performance review language压力话术 + 绩效考核语言 | Cognitive framework + structured methodology + scene routing认知框架 + 结构化方法论 + 场景路由 |
| Behavioral Constraints行为约束 | 7-item checklist7 项检查清单 | 5 Directives + 11 Anti-Patterns + 4 Dojo Quality Gates + Verification Matrix五敕令 + 反模式十一戒 + 四道场质量门 + 验证矩阵 |
| Failure Handling失败处理 | 4-level pressure escalation4 级压力升级 | 6-Stage Battle Momentum + Battle Domain Card + 9-Directive Audit + 3-Tier Stop-Loss六阶战势 + 战域卡 + 九令洞鉴 + 止损三阶 |
| Scene Adaptation场景适配 | Generic debugging通用调试 | 9 scenes × 5 cognitive formations × 16 wisdom sources9 场景 × 5 认知阵 × 16 思想源动态组合 |
| Thinking Transparency思维透明 | None无 | 5 Resonance Forms: Chain · Evidence · Tree · Mind · Contract共振五式：明链·明证·明树·明心·明约 |
| Multi-Agent | None无 | Leader / Teammate / Coach — separation of powersLeader/Teammate/Coach 三权分立 |
| User Experience用户体验 | Cold performance review rhetoric冷硬绩效话术 | Eastern wisdom + spirit animal totems — warm professional collaboration东方智慧 + 灵兽图腾，有温度的专业协作 |
| Benchmark | Composite -- | Composite -- |
One-click install, auto-detects installed platforms 一键安装，自动检测已安装平台
One command, interactive platform and language selection 一条命令，交互式选择平台和语言
claude install-plugin https://github.com/share-skills/pi
Claude Code 用户：也可以通过 Marketplace 安装：claude install-plugin https://github.com/share-skills/pi
For large models: Claude Opus/Sonnet, GPT-4o, Gemini Pro. Full cognitive framework, all capabilities. 适合大模型：Claude Opus/Sonnet、GPT-4o、Gemini Pro。完整认知框架，全部能力。
Lightweight bootstrap + on-demand loading. Starts lean, progressively loads capabilities as scenes activate. Ideal for context-sensitive workflows. 轻量引导 + 按需加载。启动精简，场景激活时渐进式加载能力模块。适合上下文敏感的工作流。
Choose language during install. Both versions are behaviorally equivalent with aligned triggers. 安装时选择语言。中英文行为等价，结构一致，触发条件对齐。