MoonshotAI
Featured
Kimi-K3
Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides a concrete, field-tested operational framework designed specifically for weak models, with explicit rules, judgment rubrics, and a bootstrap diagnostics phase that addresses the environment gap between template creation and deployment.
Who it is for
- Developers using Claude Code or similar harnesses with CLAUDE.md + subagent mechanisms
- Teams wanting to delegate project management to lower-cost (Sonnet/Haiku) models
- Engineers building reproducible AI-assisted project pipelines with explicit failure modes
Use cases
- Automating project onboarding and task dispatch with a low-cost model
- Enforcing quality gates and escalation rules in AI-driven code reviews
- Storing and evolving team lessons in a structured, model-parseable format
- Replacing ad-hoc prompt engineering with a standardized ops/ library
Strengths
- Designed for low-tier models: all rules are concrete, with positive/negative examples and judgment criteria
- Self-contained routing: CLAUDE.md stays under 60 lines, details live in ops/ for maintainability
- Evolutionary: lessons are captured and periodically pruned per maintenance protocol
- Honest about limitations: explicitly documents when to upgrade model, seek external opinion, or admit inability
Considerations
- Cannot solve ambiguous problem definition or taste-based decisions – requires model upgrade or external input
- Template assumes a specific CLAUDE.md + subagent harness environment; initial bootstrap diagnostics are needed to adapt to actual project contexts
- Model IDs in templates are snapshots from 2026-07 and may need updates for newer model versions
README quick start
project-ops-template：可移植的模型調度制度包
一套讓較低階模型（Sonnet / Haiku 等級）能穩定接管專案的制度檔案。適用於 Claude Code 或任何有 CLAUDE.md + subagent 機制的 harness。
包內容
project-ops-template/
├── README.md ← 本檔（部署後可刪）
├── CLAUDE.md.template ← 精簡路由型 CLAUDE.md 骨架
└── ops/
├── 00-bootstrap.md ← 首次進駐診斷：環境盤點 + 三大漏洞自檢（A）
├── 10-dispatch.md ← 模型調度守則：派工/升降級/驗證不自驗（C）
├── 20-judgment.md ← 判斷力 rubric：升級/完成/問人/換路/品質底線（D）
├── 30-prompt-templates.md ← 五類派工模板：搜尋/實作/重構/研究/審查（E）
├── 40-maintenance.md ← 維護協議：誰能改什麼、教訓格式、精簡週期（F）
├── 50-letter.md ← 給未來 session 的信：三件事 + 退化預防（G）
└── lessons.md ← 踩坑記錄（空模板）
部署三步（在目標專案的第一個 session 執行）
部署步驟請看：DEPLOY.md
設計原則（改動本制度前先理解）
- CLAUDE.md 是路由不是百科：60 行以內，細節全在 ops/。
- 弱模型需要明確：所有規則具體到可執行、附判準與正反例；抽象要求視同沒寫。
- 不依賴高階能力：全部流程 Sonnet 等級可跑；超出範圍的（模糊題、品味判斷）在
20-judgment.md第 6 節明寫了退場方式。 - 可長期演化：教訓進 lessons.md、按
40-maintenance.md收編與精簡，防止制度肥大或被悄悄放寬。
已知極限（誠實條款）
本制度補得了執行品質（拆解、驗收、多樣本評審），補不了模糊題的問題定義與品味判斷——遇到時的處置寫在 ops/20-judgment.md 第 6 節：升級模型、外部第二意見、或明說做不到。此外，本模板由不在目標環境內的 session 產出，00-bootstrap.md 的診斷程序就是為此設計的：所有環境相關事實以實地盤點為準，模板內的模型 id 僅為 2026-07 的參考值。
Related repositories
Similar projects matched by category, topics, and programming language.
xuchonglang
Featured
investing-for-beginners
A structured investing guide for Chinese beginners covering US stocks, options, and cryptocurrency, with focus on foundational concepts and risk awareness.
Krishnagangwal
Featured
CS-Fundamentals
A curated collection of Computer Science fundamentals (PDFs, notes, cheatsheets, interview question banks) for placement preparation, covering seven core subjects plus general resources.