A single open entry point to 187 governance-oriented agent skill packages, organised across six practice lines. Every package lists its AI-protocol status, executability, and a public link wherever a verified remote release exists.
面向治理学的 Agent 技能生态统一入口：187 个技能包按六条业务线编排。每个包标注 AI 协议达标与可执行状态；凡已核对的公开发布，均给出公开入口链接。
The registry consolidates 187 skill packages across six practice lines. All 187 carry a machine-readable SKILL.md contract (when-to-use / usage / workflow); 142 ship runnable scripts. 152 packages are already published on a public ModelScope remote; the rest are staged locally and marked pending release. Counts below are computed from the source registry, not estimated.
本索引整合六条业务线上的 187 个技能包。全部 187 个均带机器可读的 SKILL.md 协议（何时用 / 用法 / 流程）；142 个含可执行脚本。152 个包已发布至公开 ModelScope 远端，其余本地暂存、待发布。以下数量均由来源注册表实际统计得出。
Filter by practice line, AI-readiness or executability, or search by package name, slug or ID. "Public entry" links only where a remote URL was verified in the source index; staged packages show a neutral pending-release badge instead of a guessed link.
可按业务线、AI 协议达标状态或可执行状态筛选，也可按包名、slug、ID 关键词搜索。「公开入口」仅在来源索引核对过远端 URL 时给出；本地暂存的包显示中性的「待发布」标记，不推测链接。
| ID | Package | Practice line | AI protocol | Executable | 评测 | Public entry |
|---|
The entries below are flagged "public-ready" in the source public-asset index (2026-10-16). Each card lists the verified remote URL where one exists, its license and the reproducible command. Entries without a verified remote URL are shown as staged pending registration — no URL is invented.
下列条目来自来源《公开资产索引》（2026-10-16）中标记「公开就绪」的资产。每张卡片给出已核对的远端 URL、许可证与可复现命令；未核对到远端 URL 的条目标为「远程地址待登记」，不编造链接。
Methodology collections: one repo, one lifecycle narrative, sub-skills referenced by slug (never vendored). LGD three-laws badges below are read from the live dashboard (2026-10-17): registered 113/113 · evidenced 112/113 · gated 112/113.
方法论合集：单仓、一条生命周期叙事，子技能按 slug 引用（不复制实现）。下列 LGD 三律徽标实读自 2026-10-17 看板：有籍 113/113 · 有证 112/113 · 有门禁 112/113。
One operating system for the full governance lifecycle: SPEC -> PLAN -> GATE -> AUDIT. 9 sub-skills orchestrated by slug (red-team, injection shield, leak scan, desensitization, kill-switch, badge issue/verify, skill & publish gates). 788 cumulative ModelScope downloads across the 9; batch eval 7/7 executed, FAIL 0.
一套操作系统跑通治理全生命周期：SPEC（定界）-> PLAN（防御纵深）-> GATE（两道硬门）-> AUDIT（凭证留痕）。按 slug 编排 9 个子技能。9 子技能累计 788 次 ModelScope 下载；批量评测实跑 7/7、FAIL 0。
有籍 113/113 ✓ · 有证 112/113 ✓ · 有门禁 112/113 ✓ · GitHub repo · SkillPie · ClawHub
One install that answers who your agent is, remembers everything across sessions, verifies its identity, and registers every output: REGISTER -> REMEMBER -> VERIFY -> OUTPUT -> EVOLVE. 8 sub-skills orchestrated by slug (boarding-pass, memory-keeper, session-continuity, context-engineering, verified-agent-identity flagship, loop-guard, output-registry, evolution). 1,013 cumulative ModelScope downloads across the 8; all 8 three-laws green (8/8).
一次安装让智能体知道自己是谁、记住跨会话一切、可核验身份、并登记每次产出：登记->记忆->验证->输出->进化。按 slug 编排 8 个子技能（登机牌/长期记忆/跨会话连续性/上下文工程/身份核验旗舰/失控护栏/产出登记/进化飞轮）。8 子技能累计 1,013 次 ModelScope 下载；8 包三律全绿 8/8。
有籍 8/8 ✓ · 有证 8/8 ✓ · 有门禁 8/8 ✓ · GitHub repo · SkillPie (pending CLI) · ClawHub (pending auth)
One install, your agent stops leaking secrets and obeying injected instructions: RED-TEAM -> SHIELD -> LEAK -> GUARD -> GATE. 9 sub-skills orchestrated by slug (red-team kit [confidential, name only], injection drill/shield, leak scanner, token-disclosure check, tool-call guard, loop guard, MCP security scan, skill admission gate). 718 cumulative ModelScope downloads across the 9; batch eval RPT-20261016-TOP20-B: 9/9 executable PASS, 9/9 reproducible PASS, FAIL 0.
一次安装，让智能体不再泄漏密钥、不再被注入指令带偏：红队演练->注入防护->泄漏扫描->运行时守卫->准入门禁。按 slug 编排 9 个子技能（红队包[保密仅引用能力名]/注入演练与防护/泄漏扫描/披露检查/工具调用守卫/循环护栏/MCP 安全扫描/技能准入门）。9 子技能累计 718 次 ModelScope 下载；批量评测 RPT-20261016-TOP20-B：9/9 可执行 PASS、9/9 可复现 PASS、FAIL 0。
有籍 9/9 ✓ · 有证 9/9 ✓ · 有门禁 9/9 ✓ · GitHub repo · SkillPie · ClawHub
Sources: (1) the organisation skill-package registry v1.0 dated 2026-10-16 (187 rows, SK-001 through SK-187); (2) the public-asset index dated 2026-10-16 (24 entries enumerated as public-ready); (3) the on-disk inventory of public ModelScope clones (154 repositories) used to decide which packages carry a verified remote link. AI-ready means the SKILL.md frontmatter contract is complete; executable means the package ships runnable scripts.
Desensitisation: no local absolute paths, credentials or internal codenames appear on this page. Two packages whose verified remote URL contains an internal codename are displayed under a neutral package name and marked pending release, so that no codename leaks through the link target. Note: the source index's summary line reports 20 public-ready entries, while its own enumeration lists 24 — this page follows the enumeration.
数据来源：(1) 组织技能包注册表 v1.0（2026-10-16，187 行，SK-001 至 SK-187）；(2) 公开资产索引（2026-10-16，枚举「公开就绪」24 条）；(3) 公开 ModelScope 克隆的本地清单（154 个仓），用于判定哪些包带已核对的远端链接。AI 协议达标指 SKILL.md frontmatter 契约完整；可执行指包内含可运行脚本。
脱敏说明：本页不含任何本地绝对路径、凭据或内部代号。两个已核对远端 URL 中含内部代号的包，以中性包名展示并标注待发布，避免链接目标泄漏代号。另注：来源索引摘要行写「公开就绪 20 条」，但其枚举清单实际列 24 条；本页按枚举清单呈现。