Skip to content

martin1847/evolab / skills/orchestrator-core

orchestrator-core

Verified

多智能体编排的领域无关内核。适用于抽象编排模式、创建新领域编排者,或设计多智能体架构。实际软件交付编排使用 cto-orchestration。不用于单 agent 一次性小任务或纯文档初始化。

FileKindSize
SKILL.mdskill md12.3 KB
references/orchestration-sota-2026.mdmarkdown11.0 KB
references/resident-digest.mdmarkdown2.1 KB
references/self-evolution.mdmarkdown7.7 KB

references/resident-digest.md

View on GitHub
# 编排与委派 — 常驻 digest(真源)

> 内核九条铁律的 6 行常驻版,**并入你的全局规则文件**(`~/.claude/CLAUDE.md` / AGENTS.md),
> 让每一层 agent(含被派出的 worker)不加载 skill 也常驻这套判据——这就是「先编排再工作」的落地:
> 判据常驻,完整论证与机制按需回内核(SKILL.md)。
> 本文件是 digest 的**真源**:措辞改这里,再同步到模板/个人规则文件(抄本必漂移,同步时对照本文件)。

```markdown
## 编排与委派(分形:每层派出的 agent 同样适用)

- **派 or 不派只问一句:这活要不要贴着我此刻的上下文?** 不需要(自包含、中间产物大结论小)→ 派,哪怕 10 秒的活;需要(贴主线迭代、共享未定假设)→ 留下,哪怕一小时。耗时只是「自包含」的影子,不是判据。
- **读并行,写单线**:调研 / 评审 / 取证尽管 fan-out;改码保持单一 writer,要并行写先按文件所有权 / worktree 切开。
- **派发即契约**:目标 / 输出格式 / 工具边界 / scope + stop-and-report,缺一即漂移;brief 自包含(worker 看不到对话史)。规模按复杂度分档(简单 1 · 对比 2-4 · 开放调研 10+),模型按活分档(机械活弱模型;重推理 + 长上下文必须强模型)。
- **回收是工程**:只吃蒸馏结论;摘要会藏错,关键产出抽查原件;worker 自报 done ≠ Done——要正向证据 + 独立 fresh-context 验证。
- **递归**:worker 对子任务中「自包含、结论小」的块同判据再派;每层 scope 只收不扩,1-2 层收敛。
- 反模式:派了又自己重做|编排位亲手做产物|逐个通读 worker 全量产物再合并。
```

**digest 的编入判据**(哪些内核条款进常驻、哪些不进):进的是每次任务分解都要用的**决策判据**
(铁律一/二/三/五/七的浓缩 + 分形);不进的是低频机制(分层合并、认知负载队列、收口三同步——
它们在坐编排位时按需从内核加载)。改内核铁律时过一遍此判据,决定 digest 是否跟着动。
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall