Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add rules/sieveai-dev/sieve/cursorrulesgit clone --depth 1 https://github.com/SieveAI-dev/sieveWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.02862 | $0.02862 |
| Opus 5 | $0.01431 | $0.01431 |
| Sonnet 5 | $0.00572 | $0.00572 |
| Haiku 4.5 | $0.00286 | $0.00286 |
Grade A, and why
cursorrules scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 160 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sieve · Cursor 项目规则
本文件是 Sieve 项目的 Cursor 工作规则,分两层:文档规则(沿用用户级
.cursorrules模板骨架)+ Sieve 工程硬约束(项目级共识)。当前项目状态:早期预览(0.1.0-alpha),所有约束按生产语境强制执行。
一、文档规则
1.1 核心规则
- 改代码前必须先读
@docs/中相关文档(需求 → 设计 → API),不盲目重构 - 完成代码修改后必须更新所有关联文档,禁止单独更新一个文档而忽略关联文档
- 临时文档使用前缀
_temp-或_draft-,存放于docs/_temp/,任务完成后 24 小时内删除或归档 - 所有跨文档引用一律使用相对路径链接
1.2 文档位置约定
README.md # 唯一放在根目录的文档(项目入口)
.cursorrules # 本文件
docs/
├── design/ # 设计文档(architecture.md / data-model.md)
├── api/ # API 文档(api-reference.md)
├── specs/ # 功能技术规格(SPEC-NNN-*.md,含 SPEC-005 IPC 权威源)
├── guides/ # 使用指南(development.md / deployment.md)
├── changelog/ # 变更日志(CHANGELOG.md)
├── research/ # 调研材料
└── _temp/ # 临时文档(用完即删)
1.3 AI 行为约定
- 讨论技术方案时,主动引用
@docs/design/ - 修改 API 时,必须引用
@docs/api/ - 改 IPC 字段时,必须先改
@docs/specs/SPEC-005-ipc-protocol.md - 每次代码变更后,列出"需要更新的文档清单"
1.4 文档关联图
README.md
↓ 链接到
docs/design/architecture.md
↓ 约束
docs/api/api-reference.md + docs/specs/
↓ 记录变更
docs/changelog/CHANGELOG.md
规则:上游文档变更时,必须检查并更新所有下游文档。
1.5 文档化触发条件
| 场景 | 目标文档 | 优先级 |
|---|---|---|
| 新增功能 | 设计文档 + README + CHANGELOG | P0 |
| 修改 API(Anthropic / Unified Message) | api-reference.md + CHANGELOG |
P0 |
| 架构变更(pipeline / 模块边界) | architecture.md + CHANGELOG |
P0 |
| Bug 修复涉及检测规则逻辑变更 | CHANGELOG + 相关设计文档 | P1 |
配置变更(config.toml schema) |
deployment.md / development.md |
P1 |
| 依赖升级(含 vectorscan / rustls) | CHANGELOG | P2 |
无需文档化:纯代码格式化、注释优化、测试补充(无功能变更)。
二、Sieve 工程硬约束(十六条)
每条都是"做错就死",不是优化项,不接受任何阶段性妥协:
- Rust 栈非选项——Go regexp 慢 1000 倍,主路径 hot loop 不允许引入非 Rust 二进制依赖
- 绝不做联网 verifier——发送 token 到外部验证 = 摧毁产品定位,校验位 / entropy 全部本地完成
- fail-closed High-Risk Tool Policy Gate + 强制确认——YOLO mode 不允许关闭,签名 / shell / 敏感路径所有 Critical 工具调用必须走人工确认
- BIP39 必须做 SHA-256 checksum 验证——这是 Sieve 的差异化点,仅词表匹配不足以定级 Critical
- SSE 边界处理写大量 fuzz test——半行 chunk、跨 chunk 分隔符、嵌入 C0 控制字符、多 event 粘包、提前断流必须全部覆盖,PR 不带 fuzz 不合并
- Sieve 自身的供应链必须 sigstore + reproducible build + pinned dependencies——LiteLLM 事件就是先例。这件事比检测精度重要
- Critical 拦截 FP 必须 < 0.5%——超过即用户禁用产品
- Critical 在所有版本(包括降级模式)不可关闭——产品安全承诺,不是用户偏好
- 适配四家 agent:Claude Code / OpenClaw / Hermes / Codex CLI,UnifiedMessage 真实双协议——Anthropic Messages API + OpenAI Chat Completions 都跑通;Codex CLI 走原生 PreToolUse hook 经 IPC judge_tool_call 取裁决,OpenClaw / Hermes 两家 hook 作 UX 层、安全不变量由网关 inbound_hold 兜底 fail-closed;Gemini / Mistral 等其他协议推 Phase 2,真有用户主动要再做
- GA 时一次性公开 repo + 代码 + 文档——sigstore + reproducible build pipeline 照常跑通
- 不在 Anthropic API 协议层撒谎——不伪造 tool_use / stop_reason / id / usage / type;拦截发生时允许截 SSE 流注入
sieve_blockedevent(Sieve 自报,不是冒充模型);keep-alive comment 行不算伪造 - 不装本地 CA 做 MITM——Network Extension / 本地 CA 注入 / 系统 proxy 修改推 Phase 3 选购,Phase 1/2 不做
- 出站脱敏不打断工作流——OUT-01~05/12 高频类自动脱敏 + 状态栏 5s 通知,不弹窗;弹窗次数过多用户直接禁用产品
- 用户规则系统 fail-safe(v2.0 新增)——用户规则文件加载失败 / pattern 编译失败 / 安全 lint 拒绝 → daemon 必须正常启动 + 系统规则全功能;用户规则不能 override 或 suppress 系统 Critical;用户规则只能 High Ask/Warn/Mark,不能 Block / HookTerminal。详见 LayeredEngine 合并顺序
- 行为序列检测的保守起步 + 默认关闭(v2.0 新增)——IN-SEQ-* 仅触发 StatusBar 通知,不引入新 Block 路径;默认关闭(
[features] sequence_detection = false),用户主动 opt-in。升级为 Block 类需充分样本积累 + FP < 0.5% + 新决策记录。 - 所有入站能力必须经过 content-type 路由矩阵测试(v2.0 新增)——任何新增入站功能必须有集成测试覆盖 4 类组合(Anthropic SSE / Anthropic JSON / OpenAI SSE / OpenAI stream=false JSON);新功能只挂 SSE 不挂 JSON 视为 P0 漏洞。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 160 lines · 2,862 tokens per session scan A 576d0eaf1bce
cursorrules is a cursor rule published in the GitHub repository SieveAI-dev/sieve (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 2,862 tokens to every session, about $0.0143 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other cursor rules, from other repositories
kraken-net
Conventions for using Kraken.Net when working with Kraken Spot and Futures cryptocurrency exchange APIs in C#/.NET. Apply when generating code that interacts with Kraken API.
agent-guardrails-enforce
Before telling the user a task is done, you MUST.
selectools-development
Selectools development conventions, testing requirements, and release process.
selectools-providers
Rules for working with LLM provider adapters in selectools.
cursorrules
When the user asks about MiCA, crypto compliance, CASP, stablecoin, ART, EMT, EU crypto, use mica-crypto-mcp tools.
titanclip-monorepo
TitanClip pnpm monorepo layout, scripts, and scoping.