Skip to content

northseadl/norix-skills / agent-task-orchestration

agent-task-orchestration

Verified

Task decomposition and multi-agent orchestration with retry, checkpoint recovery, and real-time monitoring. Mixed Codex/Claude Code engine. Parallel/sequential execution.

evals/evals.json

View on GitHub
{
    "skill_name": "agent-task-orchestration",
    "evals": [
        {
            "id": 1,
            "query": "帮我把这个 PRD 拆成可执行的任务,用 Codex 并行开发",
            "should_trigger": true,
            "rationale": "Direct explicit request for task decomposition + Codex dispatch"
        },
        {
            "id": 2,
            "query": "I have a large feature spec in docs/feature-auth.md — can you break it down into smaller tasks and run them in parallel using multiple AI agents?",
            "should_trigger": true,
            "rationale": "Implicit task decomposition + parallel orchestration, no explicit Codex mention"
        },
        {
            "id": 3,
            "query": "用 Claude 来拆任务编排,架构设计用 Claude,具体实现用 Codex",
            "should_trigger": true,
            "rationale": "Mixed engine orchestration, mentions both engines + task decomposition"
        },
        {
            "id": 4,
            "query": "我需要一个监督模式,把需求分解给多个 Agent 执行,完成后自动评估质量",
            "should_trigger": true,
            "rationale": "Supervised mode + multi-agent execution + quality gate — core orchestration use case"
        },
        {
            "id": 5,
            "query": "dispatch.mjs 跑了一半挂了,帮我恢复一下继续跑",
            "should_trigger": true,
            "rationale": "Recovery/resume scenario — directly references dispatch tool failure"
        },
        {
            "id": 6,
            "query": "Run a long supervision loop that keeps dispatching follow-up tasks until all acceptance criteria pass",
            "should_trigger": true,
            "rationale": "Long-lived supervision loop with follow-up batches"
        },
        {
            "id": 7,
            "query": "create task specs for 5 Codex sessions from this requirements doc and run them sequentially",
            "should_trigger": true,
            "rationale": "Create task specs + dispatch — exact skill use case"
        },
        {
            "id": 8,
            "query": "把任务按 DAG 依赖关系编排,先跑基础设施再跑业务功能",
            "should_trigger": true,
            "rationale": "DAG-based task orchestration with dependency ordering"
        },
        {
            "id": 9,
            "query": "上次编排的任务 T3 和 T5 失败了,帮我只重新跑这两个,其他已完成的不要动",
            "should_trigger": true,
            "rationale": "Resume with selective retry — key v2 recovery feature"
        },
        {
            "id": 10,
            "query": "help me set up a supervised workflow where Codex handles each module independently and I review the results",
            "should_trigger": true,
            "rationale": "Supervised workflow with modular independence — core orchestration pattern"
        },
        {
            "id": 11,
            "query": "let's brainstorm about the best approach for our authentication system, get 3 different AI perspectives on it",
            "should_trigger": false,
            "rationale": "Multi-agent brainstorming → agent-brainstorm skill, not task orchestration"
        },
        {
            "id": 12,
            "query": "帮我建 3 个飞书任务,分别是前端开发、后端 API、数据库迁移,设好截止日期",
            "should_trigger": false,
            "rationale": "Creating Feishu tasks with deadlines → feishu-integration skill, not AI agent task decomposition"
        },
        {
            "id": 13,
            "query": "review this pull request and suggest improvements to the code quality, focus on the auth module changes",
            "should_trigger": false,
            "rationale": "Code review request, no task decomposition or orchestration"
        },
        {
            "id": 14,
            "query": "让几个 Agent 从不同角度讨论一下这个技术方案的优劣,帮我做个技术选型",
            "should_trigger": false,
            "rationale": "Multi-agent discussion/debate → agent-brainstorm, not orchestration"
        },
        {
            "id": 15,
            "query": "I want to deploy my Next.js app to Vercel and set up CI/CD with GitHub Actions",
            "should_trigger": false,
            "rationale": "Deployment + CI/CD task, not PRD decomposition or multi-agent orchestration"
        },
        {
            "id": 16,
            "query": "generate a Mermaid flowchart showing the task dependency tree for sprint 4",
            "should_trigger": false,
            "rationale": "Mermaid diagram → pm-toolkit, the mention of 'task' is coincidental"
        },
        {
            "id": 17,
            "query": "帮我用 Codex 写一个 Python 脚本来处理 CSV 数据,就这一个任务",
            "should_trigger": false,
            "rationale": "Single Codex task, explicitly says 'just this one task' — should use Codex directly"
        },
        {
            "id": 18,
            "query": "读取 Coding.net 上的 CI 构建日志,看看最近一次构建是否通过",
            "should_trigger": false,
            "rationale": "Coding.net CI → coding-net skill, mentions build but not task orchestration"
        },
        {
            "id": 19,
            "query": "我要设计一个前端页面原型,包括登录、注册、仪表盘三个页面的交互稿",
            "should_trigger": false,
            "rationale": "Front-end design/prototyping → agent-front-design skill, not task orchestration"
        },
        {
            "id": 20,
            "query": "长时间调度模式,在后台持续监控 Agent 执行进度,自动处理失败任务并恢复",
            "should_trigger": true,
            "rationale": "Long-running supervision + auto-retry + resume — exact trigger phrase combination"
        }
    ]
}
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