multi-agent-patterns

This skill should be used when the user asks to "design multi-agent system", "implement supervisor pattern", "create swarm architecture", "coordinate multiple agents", or mentions multi-agent patterns, context isolation, agent handoffs, sub-agents, or parallel agent execution.

Install with Tessl CLI

npx tessl i github:muratcankoylan/Agent-Skills-for-Context-Engineering --skill multi-agent-patterns

What are skills?

1.34x

Quality

47%

Does it follow best practices?

Impact

94%

1.34x

Average score across 3 eval scenarios

Optimize this skill with Tessl

npx tessl skill review --optimize ./skills/multi-agent-patterns/SKILL.md

SKILL.md

Review

Evals

Evaluation results

95%

Customer Support Routing System

Supervisor direct pass-through and bottleneck mitigation

Criteria

Without context

With context

Direct pass-through mechanism

100%

Pass-through used conditionally

100%

Worker output schema constraint

100%

Supervisor checkpoint/state persistence

50%

Supervisor delegates routing only

100%

Multiple specialist workers

100%

Routing logic present

100%

design_notes.md explains fidelity fix

100%

design_notes.md explains bottleneck fix

100%

Demo produces output

100%

No large files left

100%

Without context: $1.3770 · 5m 20s · 39 turns · 88 in / 19,654 out tokens

With context: $1.0612 · 3m 55s · 28 turns · 5,619 in / 13,644 out tokens

94%

63%

Investment Strategy Review Panel

Multi-agent consensus without sycophancy

Criteria

Without context

With context

No simple majority voting

100%

Weighted votes by confidence

100%

Weighted votes by expertise/domain

100%

Debate rounds with critique

58%

100%

Sycophancy detection

70%

Stall detection

62%

Process log shows debate

60%

100%

Process log shows weighted outcome

100%

At least three specialist agents

100%

No large files left

100%

Without context: $0.8919 · 4m 8s · 21 turns · 21 in / 17,551 out tokens

With context: $1.7759 · 5m 22s · 39 turns · 38 in / 21,695 out tokens

94%

10%

Document Analysis Pipeline Design

Context isolation design and failure resilience

Criteria

Without context

With context

Context isolation as primary rationale

41%

91%

Pattern choice justified by coordination needs

75%

87%

File system or minimal-context shared state

91%

Explicit handoff protocol

90%

Output validation before downstream use

60%

90%

Circuit breaker or retry with limit

91%

TTL or timeout per agent

100%

Failure demo continues pipeline

100%

No large files left

100%

Architecture addresses state bloat

100%

Without context: $0.4620 · 2m 24s · 16 turns · 15 in / 8,723 out tokens

With context: $0.5552 · 2m 38s · 17 turns · 18 in / 8,631 out tokens

Evaluated: 5 days ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Customer Support Routing System Investment Strategy Review Panel Document Analysis Pipeline Design

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.