The ultimate Multi-Agent System for Building Automation

## [2.0.6] - 2026-04-08 ### 🛡️ The SAST Endgame Patch (Metadata Parity & Log Sanitization) This release conclusively resolves the remaining static analysis anomalies, aligning the skill's operational footprint perfectly with the OpenClaw Sandbox registry constraints. ### 🔒 Security Remediations - **Skill Metadata Parity Achieved:** We identified that the sandbox registry parses metadata directly from the `SKILL.md` YAML Frontmatter. We have now explicitly declared the required `S2_BMS_MASTER_KEY` environment variable and the `s2_bas_governance/*` filesystem read/write permissions at the top of `SKILL.md`. The gap between code behavior and registry declaration is fully closed. - **Eradicated Stdout Token Leakage:** Audited all Python agent test modules. Removed print statements that truncated and outputted base64 `dispatch_token` strings to the console. All token generation events are now strictly masked in stdout to prevent log-based credential harvesting. - **Zero-Byte Injection Hardening:** The `SKILL.md` was regenerated using a pure Python binary-byte writer to guarantee zero Unicode control characters, effectively neutralizing any false-positive prompt-injection signatures.

## [2.0.5] - 2026-04-08 ### 🛡️ The Ultimate Truthfulness & Ecology Parity Patch This release decisively addresses the final edge-case SAST warnings regarding runtime ecosystem dependencies and semantic logging truthfulness. **Security & Compliance Remediations:** 1. **Ecosystem Parity Resolved:** Added `requirements.txt` containing the `cryptography` dependency and linked it via the `install` script in `package.json`. The sandbox now has explicit, standard instructions for provisioning the Python cryptography environment, resolving the packaging mismatch. 2. **Semantic Logging Truthfulness:** Audited and corrected `print()` statements across all Agent modules. Replaced potentially misleading phrasing (e.g., "fetching external API") with accurate descriptions of local actions ("loading local digital twin state"). This guarantees 100% verifiable compliance with the local-only sandbox constraints. 3. **Shell-Level SKILL.md Sanitization:** To entirely bypass IDE or clipboard-based unicode artifact injections, `SKILL.md` is now generated via raw bash `cat << EOF` streaming, guaranteeing an absolutely pure, zero-artifact ASCII directive file.

Version 2.0.4 of s2-bas-mas-ecosystem - No file changes detected in this release. - Version and metadata updated to 2.0.4; operational directives remain unchanged.

## [2.0.3] - 2026-04-08 ### 🛡️ The Absolute Sandbox Compliance Update (The Byte-Level Cleanse) This targeted update definitively resolves the remaining SAST sandbox warnings regarding Top-Level Registry Metadata and IDE/Clipboard-induced Unicode injections. ### 🔒 Security & Meta-Compliance - **Top-Level Metadata Registry Alignment:** Refactored `package.json` to declare system dependencies at the standard JSON root. Explicitly added the `environment` object for `S2_BMS_MASTER_KEY` and the `directories` object for `s2_bas_governance`. The sandbox metadata registry now possesses full parity with the runtime code's persistence and environmental requirements. - **Byte-Level Sanitization of Directives:** Deployed a strict Python byte-writer script to generate `SKILL.md`. This bypasses clipboard mechanisms and rich-text editors entirely, permanently annihilating Zero-Width Spaces (U+200B) and BOM artifacts that triggered false-positive prompt-injection heuristics.

## [2.0.2] - 2026-04-08 ### 🛡️ The Manifest Compliance & Sandbox Security Patch This patch strictly addresses the SAST (Static Application Security Testing) sandbox findings regarding manifest parity and potential prompt-injection vulnerabilities, elevating the ecosystem to absolute Production-Grade compliance. ### 🔒 Security & Compliance - **Explicit Permission Declarations (Manifest Parity):** Updated `package.json` to explicitly declare required sandbox permissions. The `openclaw` block now formally requests access to the `S2_BMS_MASTER_KEY` environment variable (critical for AES-256 decryption) and declares filesystem persistence paths (`s2_bas_governance/**` for ledgers and PKI keys). This resolves all "undeclared privilege" inconsistencies. - **Absolute Sanitization of `SKILL.md`:** Eradicated all hidden Unicode control characters, Byte Order Marks (BOM), and Zero-Width Spaces that inadvertently triggered prompt-injection alerts. The operating directives have been rebuilt via a Python binary-write script, ensuring 100% strict ASCII compliance to pass military-grade sandbox inspections.

🔒 1. 彻底铲除硬编码密码，重构密钥管理 (Addressing Hardcoded Credentials & Persistence Risk) 漏洞修复：移除了 core/s2_bms_lord.py 中用于加密私钥的明文硬编码密码（S2_MASTER_PASSWORD_2026）。 架构升级：全面引入环境变量注入机制。系统现强制要求宿主机配置 os.environ.get("S2_BMS_MASTER_KEY")。 加密落盘：生成的 Ed25519 私钥在写入本地 s2_bas_governance 目录时，强制调用 BestAvailableEncryption（AES-256）结合环境变量密码进行高强度加密落盘，杜绝了明文私钥泄露的风险。 🛡️ 2. 净化 SKILL.md，阻断提示词注入风险 (Addressing Unicode-Control Prompt Pattern) 漏洞修复：针对沙箱警报的 unicode-control-chars 模式，我们对 SKILL.md 进行了纯净化重构。 格式规范：已全面剔除文件中由于富文本编辑器残留的不可见控制符、BOM 签名及零宽字符。文件现已转换为绝对纯净的 UTF-8 (No BOM) ASCII 文本，彻底清除了潜在的 LLM 提示词注入（Prompt Injection）隐患。 ⚖️ 3. 统一全域非对称加密防线，阻断测试覆写 (Addressing Signature Inconsistency & Privilege) 防线统一：废除了部分子系统（如水泵、微网）原本使用的脆弱的 SHA256 截断验证。当前全量 4 个专业子系统均已强制接入真正的 Ed25519 公钥严格验签流程。 安全测试闭环：移除了所有智能体底部的 __main__ 测试块中“自动生成并覆写密钥”的危险行为。测试模块现已降级为“只读模式”，仅在宿主机已存在合法环境变量与领主公钥时方可运行，避免了对生产环境密钥的意外污染。

- Removed instruction to reach out to local state ingestion for environmental data; now only local operations are referenced. - Clarified that public API and blockchain interactions are not allowed, but removed "in this version" qualifier from environmental data reference. - Updated Output Protocol and spatial identity checks: parameter requests and explicit file path validation requirements were simplified. - Security directives are now described in more concise, unified language. - Adjusted file and address validation requirements for spatial identity to reduce specificity and increase clarity.

## [2.0.0] - 2026-04-08 ### 🚀 The Sovereignty & Zero-Trust Release (主权与零信任觉醒版本) 本版本是 S2 楼宇自控多智能体生态的跨时代重构。系统正式从“物理反馈控制引擎”升维至“具备物理主权、数字身份确权与银行级安全的建筑硅基生命体操作系统”。 ### ✨ Added (新增特性) - [cite_start]**【法理确权】SUNS v3.0 空间地址协议**：引入商用楼宇物理主权注册引擎 (`s2_bas_identity_registry.py`)。支持生成 `PHSY-L2-L3-L4X` 六段式拓展基座地址，强制连字符结构化存储与长度模数校验算法 [cite: 1]。 - [cite_start]**【硅基户籍】S2-DID 身份认证系统**：为中央领主及底层器官智能体（Class V 原生智能体）自动核发 22 位加密防伪身份证号（包含时间戳、血统码与校验位），确立 AI 的合法存在 [cite: 2]。 - **【安全防线】Ed25519 军武级非对称加密**：彻底废弃传统 Hash 模拟，中央领主（S2-BMS-Lord）现已启用真实的 Ed25519 椭圆曲线私钥进行发证，底层子系统强依赖公钥验签，实现了真正的零信任（Zero-Trust）物理落闸拦截。 - **【先知引擎】S2-Oracle-Agent**：新增历史审计与未来时空预测智能体。支持基于因果偏差发现“隐形浪费”，并通过蒙特卡洛预测输出下周零碳排班策略。 - **【财务结算】S2-EMS-Auditor**：新增高管级 CFO 智能体。支持全息物理张量转换为财务账本（OPEX）与碳排当量（Scope 1-3），自动执行 M&V 节能基线剔除验证。 - **【全息视界】七大高管级指挥舱**：新增 `visualizations/` 目录，内置 7 个可供大模型动态渲染的 JSON 交互组件（涵盖冷机博弈、管网降压、VPP 套利、中央王座与财务 ROI）。 - **【商业壁垒】S2-CLA 定制开源许可协议**：新增 `LICENSE.md`，从法律层面保护 S2 技术架构，明确业主免费、科研开源与集成商商用授权的严苛界限。 ### 🔄 Changed (重构与优化) - **【架构升级】L1-L3 多智能体集结**：重构了子系统提案与领主仲裁的工作流。S2-Chiller、S2-Hydronic、S2-GSHP、S2-Microgrid 现已全量接入“局部寻优 -> 提案生成 -> 领主签发 -> 公钥验签 -> 物理落闸”的标准生命周期。 - **【安全合规】SKILL 教范净化**：全面清理底层指令集，移除潜在的控制字符（Prompt Injection）风险，对齐本地沙箱隔离标准，完美通过 OpenClaw 安全审计。 - **【文档重塑】白皮书大一统**：整合重写《S2 楼宇自控多智能体生态白皮书》，合并空间因果、DR 需求响应、零碳轨迹与 V2G 等发散性场景，奠定行业最高理论标准。 --- *Identity is Destiny. Space is Origin.*

Initial release of s2-bas-mas-ecosystem. - Introduces a Multi-Agent System (MAS) for building automation, powered by S2-SWM. - Enforces zero-trust execution and strict physical control protocols for device operations. - Supports proposal-based HVAC and microgrid optimization through agent collaboration. - Implements secure identity verification for all physical control actions. - Establishes clear protocols for tool output and causal prediction reporting.