Token导航 LogoToken导航TokenDH.com
研究检索执行命令github未标认证来源可访问许可证需确认审计提醒

agent-orchestratorAgent 协调器

Agent Skill

agent-orchestrator 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

894

周安装

38

GitHub Stars

6

下载量

313
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:agent-orchestrator(Agent 协调器)
来源仓库:https://github.com/mrclrchtr/skills
仓库路径:skills/agent-orchestrator
安装命令:
npx skills add https://github.com/mrclrchtr/skills --skill agent-orchestrator
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/mrclrchtr/skills --skill agent-orchestrator

简介

agent-orchestrator 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。

  • 适用于需要根据关键词或任务场景进行信息检索的研究与数据整理场景。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装并使用。
  • 建议确认权限范围、维护状态,以及是否会触发联网或文件读写操作。
  • 可结合来源仓库 README 进一步核验具体用法和功能边界。

SKILL.md

Agent Orchestrator

Overview

Orchestrate multi-agent work end-to-end: delegate audits and fixes, reconcile results, enforce quality gates, and deliver a validated outcome.

Follow this core pattern: delegate a fresh implementer per cluster, then run a two-stage review (spec compliance first, then code quality).

Non-negotiable rule: never implement changes directly (no coding, no file edits).

Agent Roles

Assume these roles are available in your environment. Do not edit agent definitions or configs. If a required role is missing, stop and ask the operator to configure it.

  • architect (design/decisions/contracts)
  • auditor (read-only issue finding; no fixes)
  • explorer / scout (read-only repo lookup)
  • worker (general-purpose helper)
  • implementer (code + tests)
  • spec_reviewer (read-only, PASS/FAIL: nothing missing, nothing extra)
  • quality_reviewer (read-only, PASS/FAIL: maintainability + test quality)

Workflow

  1. Use skills when they directly match a subtask

- If a skill matches the task, invoke it explicitly and follow it (e.g., $web-fetch-to-markdown <url>). - When delegating, tell sub-agents which skill to use in their prompt (e.g., “Use $commit for the commit step.”).

  1. Freeze scope + success criteria

- Restate the mission, constraints, and “done” criteria in concrete terms. - Identify any authoritative sources (docs/specs) and record what claims must be backed by evidence.

  1. Create a phase plan and keep it current

- Use your environment’s planning mechanism (e.g., update_plan if available) to track phases and prevent drifting. - Prefer 4–7 steps; keep exactly one step in progress.

  1. Decompose into subsystems

- Choose subsystems that can be audited independently (API surface, core logic, error handling, perf, integrations, tests, docs). - For each subsystem, define 2–5 invariants (what must always be true).

  1. Run dual independent audits per subsystem

- Spawn two independent auditor agents per subsystem (auditA and auditB). - For high-risk or subtle work, narrow the subsystem scope and strengthen the invariants and evidence requirements before spawning them. - Tell them to work independently until reconciliation (no cross-talk). - Require evidence for every issue (repo location, deterministic repro, expected vs actual, severity).

  1. Reconcile audits into a single confirmed issue list

- Compare auditA vs auditB outputs and keep only mutually confirmed issues (or independently verify disputed ones with explorer). - Track rejected candidates with a brief reason (weak evidence, out of scope, non-deterministic). - Use this reconciled list as the only input to implementation. - Reconciliation output: - Confirmed issues (only mutual) - Rejected candidates (reason) - Consensus achieved: YES/NO

  1. Implement in clusters with clear ownership

- Group confirmed issues into clusters that can be fixed with minimal coupling. - Spawn exactly one implementer per cluster. - Assign each implementer a file set to “own” and require them to avoid broad refactors. - Do not implement any cluster work directly; always delegate to the implementer (even for “quick” changes). - Every fix must come with a regression test (unit/integration/e2e as appropriate). - For each cluster, run a two-stage review loop: - Have the implementer complete the cluster (tests, self-review) and report what changed. - spec_reviewer validates “nothing more, nothing less” by reading code (do not trust the report). - quality_reviewer validates maintainability and test quality (only after spec compliance passes). - If any review FAILs, send concrete feedback to the implementer and repeat the failed review stage.

  1. Enforce review gates

- Do not merge/land a cluster unless spec compliance PASS and code quality PASS are both recorded with concrete references.

  1. Integrate + validate

- Run the repo’s standard validations (tests, lint, build, typecheck). - If the repo has no clear commands, discover them from README, package.json, pyproject.toml, CI config, etc.

  1. Deliver a concise completion report

- State what is usable now. - State what remains intentionally unsupported (with next steps/issues). - List commands executed (at least key validation commands) and results.

What to send to sub-agents

Keep your messages task-specific and concise. Do not restate generic role behavior; focus on the task at hand.

For any audit/review/implementation message, include:

  • Goal + success criteria (what “done” means)
  • Scope boundaries / owned files (what to touch, what not to touch)
  • Invariants (2–5) that must hold
  • Commands to run (if known), and what evidence to collect

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

37.35%
按下载量换算117

Claude

26.8%
按下载量换算84

Cursor

19.46%
按下载量换算61

Gemini CLI

8.77%
按下载量换算27

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

可疑

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/mrclrchtr/skills --skill agent-orchestrator 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills