richfrem/agent-plugins-skills / plugins/agent-orchestration/.apm/skills/orchestrating-agent-work
orchestrating-agent-work
VerifiedInteractive orchestrator for complex engineering work. Analyzes the task, conducts a diagnostic interview to select the appropriate execution strategy (Directed Acyclic Graph versus Loop, and which Loop type), plans the artifact, and supervises execution and verification. Use for "how should I orchestrate this", "plan this workflow", "should this be a DAG or a loop", "run these jobs in parallel", "delegate this to a worker agent", or "run/resume this graph manifest". Not for single trivial edits or skill self-improvement.
references/bot-delegation-patterns.md
View on GitHub# Bot Delegation Patterns: Synchronous CLI vs. Asynchronous Remote Bots
## 1. Context & Purpose
The `orchestrating-agent-work` skill coordinates tactical workers and execution topologies (Graph DAG, Dual-Loop, Swarm). Historically, execution assumed local, synchronous CLI processes (`claude`, `agy`, `codex`) running inside isolated git worktrees via shell redirection.
Modern multi-agent architectures also incorporate **commercial autonomous agent platforms ("bots")** that operate as independent cloud or sandboxed VMs. This reference defines the orchestration contracts, boundary rules, and handoff protocols when dispatching to or receiving tasks from these platforms.
---
## 2. Platform Taxonomy & Capability Matrix
| Platform | Primary Modality | Strengths & Capabilities | Availability & Subscription | Orchestration Pattern |
|---|---|---|---|---|
| **xAI: Grok Bot** | Autonomous Cloud VM + iOS Frameworks | Outbound PSTN phone calls, disposable agent email (`@mail.grokbot.com`), 1Password/Shop Pay checkouts | Bundled in Cursor Premium ($30/mo) & X Premium+ | **Asynchronous Git / PR Handoff**: Dispatch via committed brief or webhook; results ingest via branch merge. |
| **Anthropic: Claude Duet** | Enterprise SaaS Connector | Proactive channel audit, calendar/inbox risk briefing, human-in-the-loop payment confirmation | US-only enterprise/preview beta; multi-SaaS auth | **Upstream Task Planner**: Claude Duet identifies operational issues and stages work packets for CLI loops. |
| **OpenAI: Dots (Operator)** | Sandboxed Desktop / Web VM | Direct GUI pixel-stream vision-action (mouse clicks, keyboard input), multi-step web forms lacking APIs | US-only preview rollout (ChatGPT Plus / Pro / Enterprise) | **Headless Browser Workhorse**: Dispatch task packet with target URLs and verification criteria. |
| **Google: Gemini Agent (Astra)** | Chrome Browser Engine + Multimodal | Real-time screen sight, live video synthesis, Deep Research cross-tab web dossiers | Global rollout (Google AI Studio / Advanced / AGY) | **Deep Research Layer**: Dispatches multi-source exploration jobs and writes synthesis dossiers. |
---
## 3. Synchronous CLI vs. Asynchronous Bot Execution
```
┌────────────────────────────────────────────────────────┐
│ orchestrating-agent-work │
│ (Strategic Planner / Router) │
└───────────────────┬────────────────────────────────────┘
│
┌─────────────────────────┴─────────────────────────┐
▼ ▼
┌─────────────────────────────────┐ ┌─────────────────────────────────┐
│ Synchronous In-Process CLI │ │ Asynchronous Remote Bot │
│ (Dual-Loop / Swarm / Local DAG) │ │ (Grok Bot, Dots, Duet, Astra) │
├─────────────────────────────────┤ ├─────────────────────────────────┤
│ • Local subprocess / worktree │ │ • Isolated Cloud / Web VM │
│ • Stdin redirection (< /dev/null│ │ • No direct stdio REPL pipe │
│ • Direct file read/write │ │ • PR, branch, or webhook handoff│
│ • Immediate test suite feedback │ │ • Asynchronous join barrier │
└─────────────────────────────────┘ └─────────────────────────────────┘
```
---
## 4. Orchestration Boundaries & Invariants
1. **No Stdio Blocking for Remote Bots**: Never attempt to run remote VM bots (`grok`, `dots`) as synchronous terminal pipes or blocking subprocesses in `graph_runner.py` or `agent_orchestrator.py`.
2. **Asynchronous Barrier Steps**: In a Directed Acyclic Graph (`graph-manifest.json`), external bot dispatches must be structured as:
- Node 1: `write_handoff_brief` (author atomic request packet in `handoffs/`).
- Node 2: `await_external_artifact` (barrier waiting for branch commit, PR, or human confirmation).
- Node 3: `verify_and_integrate` (run test suite and compiler checks on returned payload).
3. **Workspace Boundary Rule**:
- **Enterprise / Work Projects**: GitHub Copilot CLI is permitted.
- **Sovereign / Personal Research**: Restricted strictly to personal subscriptions (Claude Pro, ChatGPT Plus, Gemini Pro). Never route work-scoped tokens to sovereign research repositories.
4. **Anti-Asch Verification**: Remote bot outputs are untrusted until verified. Run linters, unit tests, and SHA-256 payload integrity audits before admitting changes to repository state.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 471k | Updated | Claude CodeCodex | Verified | Install |
| 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: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |