skills.rest
ruvnet
ruvnetCOMMUNITY
Explore all skills by this author →
Website𝕏@ruvGitHub

intelligence-route

Selects agent and model tier for tasks using learned routing patterns.

70.1k|8.4k|Updated Jun 2, 2025
Search Tags:#agent selection#cost optimization#pattern matching#model routing#task routing#codemod#Claude-Flow
One-click install
npx skills add https://github.com/ruvnet/claude-flow --skill intelligence-route
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: intelligence-route
Source: https://github.com/ruvnet/claude-flow/tree/main/plugins/ruflo-intelligence/skills/intelligence-route
Command: npx skills add https://github.com/ruvnet/claude-flow --skill intelligence-route

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Choosing the right agent and model tier for each task is guesswork that wastes cost on simple tasks or underpowers complex ones. This Skill replaces manual selection with data-driven routing based on learned patterns and a 3-tier model selector.

Core Features & Use Cases

  • Agent Recommendation: Calls hooks_route to get a recommended agent with confidence score and reasoning for any task description.
  • 3-Tier Model Routing: Routes tasks to deterministic codemods ($0, ~1ms), Haiku (~$0.0002), or Sonnet/Opus based on complexity.
  • Pattern Search & Prediction: Searches prior successful patterns and predicts outcomes via neural_predict before spawning agents.
  • Outcome Learning: Records task success or failure via hooks_model-outcome so the router improves over time.
  • Use Case: Before refactoring a module, route the task to discover it qualifies for a $0 deterministic codemod instead of an LLM call, then record the outcome to train future routing.

Quick Start

Ask the assistant to route the task "remove console statements from src/utils.ts" and explain why it chose that model tier.

Frequently Asked Questions about intelligence-route

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I route a task to the right model tier automatically?▼

Call hooks_route with the task description to get an agent recommendation, then hooks_model-route for the Haiku/Sonnet/Opus tier. The router returns confidence scores and reasoning for each decision.

What tasks qualify for the $0 deterministic codemod tier?▼

Structural transforms with no LLM judgment qualify: var-to-const, remove-console, and add-logging. Tasks like add-types or add-error-handling require judgment and route to Tier 2 or 3 models instead.

How does the routing system learn from past tasks?▼

After each task completes, call hooks_model-outcome with the taskId, success flag, and model used. The router trains on these outcomes; skipping outcome recording means no learning occurs.

Can I see why a specific routing decision was made?▼

Yes, pass the --why flag or call hooks_explain with the routing topic. It surfaces the rationale including confidence scores and pattern matches behind the agent and model selection.

When should I not use automated model routing?▼

Avoid it for trivial one-line tasks where routing overhead exceeds the task cost, or when you already know a task requires a specific model for compliance or consistency reasons.

Related Skills

Explore More Software Engineering Skills→
hjemmesidekongenhjemmesidekongen
1

task-decomposer

Decompose complex task descriptions into ordered subtasks with complexity estimates and dependency chains.

Community
Advanced
hiddink-aihiddink-ai
1

model-escalation

Track task failures and recommend higher-tier model usage by threshold.

Official
Intermediate
bordenetbordenet
8

measurement-integrity

Cross-validate metrics with dual methods and completeness checks before reporting.

Community
Advanced
barichterbarichter

model-selection

Select and assign models to agents using a 4-layer hierarchy with fallback strategies.

Community
Advanced