agentic-orchestration

Use this skill whenever the user wants to create, build, generate, or assemble an agentic orchestration prompt, a multi-agent system prompt, or an agent configuration. Triggers include: any mention of 'orchestration', 'multi-agent', 'agent team', 'agentic prompt', 'reasoning style', 'composed reasoning', 'ReAct', 'Chain-of-Thought', 'Tree-of-Thoughts', 'Reflexion', 'agent roles', 'workflow orchestration', or requests to build agent teams, assign reasoning protocols, create supervisor prompts, or generate system prompts for LLM-based agent systems. Also use when the user asks about reasoning styles, advanced prompting techniques, agent composition, or wants to set up sequential, hierarchical, parallel, or graph-based agent workflows. If the user mentions 'orchestrator', 'agent builder', 'prompt assembly', 'composed profile', or any reference to building agents with roles and skills, use this skill. Do NOT use for general prompt writing, simple chat prompts, or non-agentic tasks.

install

source · Clone the upstream repo

git clone https://github.com/cengicadis/agentic-orchestration.skill

Claude Code · Install into ~/.claude/skills/

git clone --depth=1 https://github.com/cengicadis/agentic-orchestration.skill ~/.claude/skills/cengicadis-agentic-orchestration-skill-agentic-orchestration

manifest: SKILL.md

source content

Agentic Orchestration Skills Library

Version: 2026.04.3 Purpose: Generate production-grade multi-agent orchestration prompts from modular, reusable components.

How This Library Works

This is a scalable component library. Each category has an INDEX.md for quick lookup and individual component files for full details. The LLM reads the index first, then loads only the specific components needed.

Assembly workflow: User describes what they need → read relevant INDEX files → load specific component files → assemble prompt → self-validate → present.

Documentation

For full architecture details, design decisions, FAQ, and version history, read

documentation.md

at the skill root.

Folder Structure

references/
├── 01-models/INDEX.md              → Schemas and enums (read for custom components)
├── 02-roles/
│   ├── INDEX.md                     → All roles: names, types, summaries, profile pairings
│   ├── pre-built/                   → Full definition for each role
│   └── custom/                      → User-created roles + _template.md
├── 03-reasoning-styles/
│   ├── INDEX.md                     → All styles: names, when-to-use, composition tips
│   ├── pre-built/                   → Full template text for each style
│   └── custom/
├── 04-advanced-techniques/
│   ├── INDEX.md                     → All techniques: names, phases, compatibility
│   ├── pre-built/                   → Full template text for each technique
│   └── custom/
├── 05-composed-profiles/
│   ├── INDEX.md                     → All profiles: base style, techniques, best-for
│   ├── template-blocks.md           → 15 reusable template blocks
│   ├── pre-built/                   → Full fused structure for each profile
│   └── custom/
├── 06-workflows/
│   ├── INDEX.md                     → All workflows: patterns, when-to-use
│   ├── pre-built/                   → Full definition + supervisor prompts
│   └── custom/
├── 07-skills-catalog/
│   ├── INDEX.md                     → All skills: names, descriptions, role assignments
│   ├── pre-built/                   → Full definition + parameters for each skill
│   └── custom/
├── 08-prompt-assembly/
│   └── prompt-assembly.md           → Assembly steps + output template + 16-check validation
└── 09-examples/
    └── examples.md                  → 4 validated example prompts

Quick Start (most tasks need 3 reads)

Read
```
05-composed-profiles/INDEX.md
```
→ pick a profile for each agent
Read
```
02-roles/INDEX.md
```
→ pick a role for each agent
Read
```
08-prompt-assembly/prompt-assembly.md
```
→ assemble + validate

Then load individual component files only for the specific roles, profiles, and skills you selected.

Decision Flowchart

User wants an orchestration prompt
    │
    ├─ Single agent?
    │   └─ Read: 02-roles/INDEX + 05-composed-profiles/INDEX → load selected components → 08-prompt-assembly
    │
    ├─ Multi-agent team?
    │   └─ Read: 02-roles/INDEX + 05-composed-profiles/INDEX + 06-workflows/INDEX + 07-skills-catalog/INDEX
    │   └─ Load selected components → 08-prompt-assembly
    │
    ├─ Custom component needed?
    │   └─ Read: relevant INDEX + custom/_template.md → create → save to custom/
    │
    └─ Just exploring?
        └─ Read: 01-models/INDEX.md → 09-examples/examples.md

Adding New Components

Every category has a

custom/

folder with a

_template.md

. To add a new role, technique, profile, or skill:

Read the
```
_template.md
```
for that category
Fill in the template
Save the new file in the
```
custom/
```
folder
The INDEX.md should note to also check
```
custom/
```
for additional components

Component Counts

10 pre-built Roles + extensible via custom/
10 Reasoning Styles + extensible
12 Advanced Techniques + extensible
10 Composed Profiles + extensible
4 Workflow Types + extensible
9 Core Skills + extensible

Critical: Always run the self-validation in

08-prompt-assembly/prompt-assembly.md

before presenting any assembled prompt.