agentsope/SkillAlchemy / skills/agentsop-agent-topology-selection
agentsop-agent-topology-selection
VerifiedCross-framework enhancement overlay for choosing a multi-agent topology BEFORE writing any agent. A binary-question rubric — is single-agent + tools enough? do agents need to know about each other? does the output need one voice? — maps the answer to single-agent / supervisor / swarm / sequential / hierarchical. Activates when a coder agent is tempted to "split the work into roles" or reaches for a multi-agent framework. Encodes the *selection rubric* that the per-framework skills assume but never surface. Search keywords: when to use multi-agent, single vs multi agent, do I need multiple agents, supervisor vs swarm, multi-agent vs single agent, agent team design.
| File | Kind | Size |
|---|---|---|
| README.md | markdown | 3.9 KB |
| SKILL.md | skill md | 18.7 KB |
| intermediate/operation_candidates.json | config | 7.8 KB |
| references/R1-source-evidence.md | markdown | 7.0 KB |
README.md
View on GitHub# agent-topology-selection skill
Cross-framework **enhancement overlay** for **multi-agent topology selection** —
deciding *whether* to use multiple agents at all, and if so, *which* topology
(single-agent / sequential / supervisor / swarm / hierarchical) **before** writing
any agent code.
This is the **O6 gap skill** in the Phase-D enhance pass. CrewAI exposes
`sequential` / `hierarchical` and LangGraph exposes supervisor / swarm, but no
skill surfaces the *selection rubric* that should drive the choice. This is that
skill.
## The rubric (binary questions)
```
Q0 Is single-agent + tools enough? → YES: single-agent. STOP.
Q1 Do agents need to know about each other? → NO: supervisor / sequential
Q2 Must the output speak with one voice? → YES: supervisor / NO: swarm
≥6 specialists grouping into teams → hierarchical (grouping only)
```
## Core insight
> Most "multi-agent" problems are single-agent + tools. Add agents only when
> *context isolation* or *parallel expertise* genuinely demands it. The single
> agent is the baseline every topology must beat — defend it first.
## The nuance that matters
LangGraph's *own* benchmark found swarm slightly beats supervisor on accuracy and
uses fewer tokens — **yet they ship supervisor as the recommended default.** The
default encodes *third-party-agent safety* (single audit funnel), not accuracy. A
coder agent copying the default blindly leaves accuracy and tokens on the table.
The rubric forces the decision onto the two binary questions instead.
## Layout
```
d-agent-topology-selection-skill/
├── SKILL.md # 7-section SOP (activation → cross-framework table)
├── README.md # this file
├── references/
│ └── R1-source-evidence.md # every cited claim resolved to a source SKILL line
└── intermediate/
└── operation_candidates.json # raw trigger/action/output/evidence operations
```
## Contents
8 operations (defend-baseline, single-vs-multi gate, static-order gate, the two
binary gates, the supervisor-default caveat, tune-before-switch, bound-the-loop),
2 dilemma cases (swarm-beats-supervisor-but-ship-supervisor; CrewAI hierarchical is
structurally broken for routing), 6 anti-patterns, and the cross-framework mapping
(CrewAI sequential/hierarchical, LangGraph supervisor/swarm, OpenAI Swarm handoff,
single-agent + tools baseline).
## ENHANCE overlay
This is an **overlay**, not a replacement. For the per-framework API it cross-links
the base skills inline as `[[name]]`:
- `[[crewai]]` — `Process.sequential` / `Process.hierarchical`, `manager_agent`,
`allow_delegation`, the broken-hierarchical-router finding.
- `[[agentsop-langgraph]]` — supervisor / swarm / hierarchical-teams decision tree, the
multi-agent benchmark.
- `[[agentsop-bounded-loop]]` — any runtime-handoff topology must be bounded before ship.
Activate this skill for the *topology decision*; descend to the base skill for the API.
## Method
Mined directly from the source SKILLs under
`/Users/5imp1ex/Desktop/Skill-Workplace/output/`: `crewai-sop-skill`,
`langgraph-sop-skill` (frontmatter cross-checked against `~/.claude/skills/crewai`).
Every load-bearing claim carries an inline `[[source · section]]` tag resolving in
`references/R1-source-evidence.md`. No fabricated APIs.
## Position in the Phase-D inventory
- **Sibling overlays**: `d-query-routing-skill` (routes by query *kind* — composes
with this skill's routing/coordination distinction), `d-bounded-loop-skill`
(bounds the chosen topology), `d-llm-engine-selection-skill`.
- **Boundary vs `d-query-routing-skill`**: that skill picks a *handler for a query*;
this skill picks an *agent topology for a task*. Routing (control flow) is a
reason **not** to use hierarchical (see Dilemma Case 2).
- **Boundary vs `[[agentsop-bounded-loop]]`**: this skill picks the topology; bounded-loop
guarantees it terminates.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |