skills.rest
arbisoft
arbisoftOFFICIAL
Explore all skills by this author →
Website𝕏@arbisoftGitHub

agentic-engineering

Coordinate AI engineering tasks with eval-first validation and cost-aware model routing.

4|7|Updated Apr 9, 2026
Search Tags:#decomposition#risk-management#ai-engineering#model-routing#eval-first#cost-discipline
One-click install
npx skills add https://github.com/arbisoft/ai-skillforge --skill agentic-engineering-arbisoft
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentic-engineering
Source: https://github.com/arbisoft/ai-skillforge/tree/main/Claude/skills/agentic-engineering
Command: npx skills add https://github.com/arbisoft/ai-skillforge --skill agentic-engineering-arbisoft

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrate engineering work for AI agents by enabling eval-first validation, structured decomposition, and cost-aware routing to ensure quality and risk controls.

Core Features & Use Cases

  • Eval-first workflow: Define capabilities, run baselines, and verify results before full implementation.
  • Task decomposition: Break work into agent-sized, independently verifiable units with clear done criteria.
  • Model routing: Allocate tasks to Haiku, Sonnet, or Opus based on complexity to balance speed and quality.
  • Cost discipline: Track model usage, retries, and timing to manage risk and cost.

Quick Start

Decompose complex engineering tasks into agent-sized units and route work by task complexity.

Frequently Asked Questions about agentic-engineering

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

FAQPage Schema
How do I manage AI engineering costs when routing tasks across different models?▼

Manage AI engineering costs by routing tasks to Haiku, Sonnet, or Opus based on complexity. This cost-aware routing balances speed and quality while tracking usage, retries, and timing to enforce cost discipline.

What is an eval-first workflow for AI agents?▼

An eval-first workflow defines capabilities, runs baselines, and verifies results before full implementation. This ensures agents meet correctness and safety standards through continuous evals and regression checks during software development.

How do I decompose complex software projects into units for AI agents?▼

Decompose complex software projects into agent-sized, independently verifiable units with clear done criteria. This structured task decomposition allows AI to implement core components while humans review risk and quality.

Can I coordinate multi-file software engineering tasks with AI while maintaining human oversight?▼

Yes, you can coordinate multi-file software engineering tasks by using controlled flow where AI implements core components. Continuous evals and regression checks ensure human reviewers can effectively validate risk and quality.

When should I use model-tier routing for AI engineering workflows?▼

Use model-tier routing for AI engineering workflows when processing complex, multi-file projects. Allocating tasks to Haiku, Sonnet, or Opus based on complexity balances execution speed, output quality, and overall cost discipline.

What are the limitations of using AI agents for complex engineering workflows without structured decomposition?▼

Without structured decomposition, AI agents struggle to verify work independently, increasing risk and reducing quality. Complex engineering workflows require clear done criteria and eval-first validation to ensure correctness and safety.

Related Skills

Explore More Software Engineering Skills→
samymitysamymity
1

agentic-engineering

Guide agentic engineering workflows with eval-first execution and model-tier routing.

Community
Advanced
linnefromicelinnefromice

eval-harness

Evaluate Claude Code sessions with code-based, model-based, and human graders.

Community
Advanced
BelichickGillisMuskBelichickGillisMusk

tps-report

Monitor task progress and costs within AI agent workflows.

Official
Intermediate
kouisokouiso

agentic-engineering

Decompose engineering tasks into verifiable units with eval-first execution.

Community
Advanced