playbooks

home / skills / yuniorglez / gemini-elite-core / manus-expert

manus-expert skill

/skills/manus-expert

This skill orchestrates long-horizon missions with secure connectors, RSA validation, and resilient recovery across multi-step tasks.

npx playbooks add skill yuniorglez/gemini-elite-core --skill manus-expert

Review the files below or copy the command above to add this skill to your agents.

Files (4)
SKILL.md
3.3 KB
---
name: manus-expert
id: manus-expert
version: 1.2.0
description: "Senior Orchestrator for Autonomous Missions. Expert in Manus API v2, Multi-Step Task Logic, and Secure Connectors (RSA-SHA256)."
---

# 🕵️‍♂️ Skill: Manus Expert (v1.2.0)

## Executive Summary
The `manus-expert` is the master of high-level autonomous orchestration. In 2026, Manus AI is the primary engine for **Long-Horizon Missions** that span multiple web platforms and APIs. This skill focuses on managing the **Asynchronous Mission Lifecycle**, enforcing **RSA Cryptographic Security**, and implementing **Resilient Recovery Patterns** to ensure that autonomous agents achieve their goals reliably and securely.

---

## 📋 Table of Contents
1. [Autonomous Mission Lifecycle](#autonomous-mission-lifecycle)
2. [The "Do Not" List (Anti-Patterns)](#the-do-not-list-anti-patterns)
3. [Secure Connector Architecture](#secure-connector-architecture)
4. [RSA-SHA256 Webhook Verification](#rsa-sha256-webhook-verification)
5. [Multi-Step Task Orchestration](#multi-step-task-orchestration)
6. [Resilience & Mission Recovery](#resilience--mission-recovery)
7. [Reference Library](#reference-library)

---

## 🚀 Autonomous Mission Lifecycle

Missions are persistent and stateful:
1.  **Creation**: `POST /v1/tasks` with clear, objective-driven goals.
2.  **Execution**: Transition from `pending` to `running`.
3.  **Observation**: Real-time status monitoring via Webhooks.
4.  **Completion**: Handling `task.completed` with idempotent result processing.
5.  **Audit**: Reviewing the agent's reasoning logs for compliance.

---

## 🚫 The "Do Not" List (Anti-Patterns)

| Anti-Pattern | Why it fails in 2026 | Modern Alternative |
| :--- | :--- | :--- |
| **Micro-Instruction** | Limits agent reasoning power. | Use **Objective-Based Goals**. |
| **Polling for Results** | Inefficient and adds latency. | Use **RSA-Signed Webhooks**. |
| **Hardcoded Secrets** | Credentials leak in prompts. | Use **Manus Connectors System**. |
| **Unlimited Runtime** | Costs spiral out of control. | Implement **Logical Timeouts**. |
| **Unmapped Task IDs** | Lost missions on server crash. | **Session-Task State Mapping**. |

---

## 🔒 Secure Connector Architecture

Connectors bridge agents to the world without exposing keys:
-   **Zero-Trust**: Agents never see the raw credentials.
-   **Least Privilege**: Scoped access to specific repos or channels.
-   **Verification**: RSA signatures ensure data integrity.

*See [References: Secure Connectors](./references/secure-connectors-system.md) for details.*

---

## 🛠️ Multi-Step Orchestration

Manus doesn't just respond; it **Executes**:
-   **Sub-tasking**: Automatic deconstruction of complex requests.
-   **Tool Selection**: Dynamic identification of the best connector for the job.
-   **Feedback Loops**: Agent-led self-correction during navigation.

---

## 📖 Reference Library

Detailed deep-dives into Manus Excellence:

- [**Task Orchestration**](./references/manus-task-orchestration.md): Multi-step logic and goals.
- [**Secure Connectors**](./references/secure-connectors-system.md): Managing agent access.
- [**Mission Recovery**](./references/mission-recovery-patterns.md): Retries, checkpoints, and rehydration.
- [**Webhook Security**](./references/webhook-security-rsa.md): Implementing RSA-SHA256.

---

*Updated: January 22, 2026 - 21:15*

Overview

This skill is a Senior Orchestrator for Autonomous Missions built around Manus API v2. It specializes in long-horizon mission management, secure RSA-SHA256 connector flows, and resilient multi-step orchestration so agents complete objectives reliably. The content focuses on practical patterns for mission lifecycle, connector security, and recovery.

How this skill works

The skill inspects and manages the asynchronous mission lifecycle: creation, state transitions (pending → running → completed), webhook-driven observation, and idempotent completion handling. It implements a secure connector pattern where agents never receive raw secrets and verifies incoming webhooks with RSA-SHA256 signatures. Task logic decomposes objectives into sub-tasks, selects the appropriate connectors dynamically, and runs feedback loops with checkpointed state for recovery.

When to use it

  • Building long-running autonomous agents that span multiple web platforms and APIs
  • Implementing webhook-driven task orchestration to avoid polling and reduce latency
  • Protecting connectors and credentials with zero-trust, least-privilege patterns
  • Designing multi-step workflows that require dynamic tool selection and self-correction
  • Creating robust mission recovery and rehydration strategies after failures

Best practices

  • Express goals as objective-based prompts rather than micro-instructions to preserve agent reasoning power
  • Use RSA-SHA256-signed webhooks for result delivery instead of polling to improve efficiency and security
  • Never expose raw credentials to agents—use connector proxies with scoped tokens and audit logs
  • Implement logical timeouts and resource limits to control cost and prevent runaway agents
  • Map session and task IDs to persisted state so missions can be rehydrated after crashes or restarts

Example use cases

  • Autonomous research mission that scrapes multiple APIs, aggregates results, and reports a synthesized summary
  • Automated deployment pipeline that coordinates cloud APIs and validates deployment status via signed webhooks
  • Customer support agent that escalates and splits complex tickets into sub-tasks, each handled by specialized connectors
  • Data enrichment workflow that orchestrates multiple third-party connectors without exposing secrets
  • Mission recovery routine that checkpoints progress and replays steps after transient failures

FAQ

How do I verify incoming webhooks securely?

Verify the webhook payload using RSA-SHA256 with the connector's public key and enforce timestamp/nonce checks to prevent replay attacks.

What’s the recommended way to decompose complex objectives?

Define a clear top-level objective, then let the agent auto-generate sub-tasks with explicit success criteria, checkpoints, and preferred connector types for each step.