skills.rest
Oruga420
Oruga420COMMUNITY
Explore all skills by this author →
Website𝕏@Oruga420GitHub

agentic-engineering

Coordinate AI agents with eval-first execution, task decomposition, and cost-aware model routing.

Updated Mar 24, 2026
Search Tags:#decomposition#agentic#code-review#ai-agents#model-routing#cost-monitoring#eval-first
One-click install
npx skills add https://github.com/Oruga420/claude-code-skills --skill agentic-engineering-oruga420
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentic-engineering
Source: https://github.com/Oruga420/claude-code-skills/tree/main/agentic-engineering
Command: npx skills add https://github.com/Oruga420/claude-code-skills --skill agentic-engineering-oruga420

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Engineering teams need a reliable method to have AI agents perform the bulk of implementation work while ensuring quality, safety, and measurable outcomes; this Skill provides operating principles and a repeatable loop to define completion criteria, validate results, and control costs.

Core Features & Use Cases

  • Eval-first execution: Define capability and regression evaluations before implementation to measure deltas and prevent regressions.
  • Task decomposition: Break work into independently verifiable 15-minute units with clear done conditions and single dominant risks.
  • Model routing & cost discipline: Route tasks across model tiers by complexity (classification, implementation, architecture) and track model, token estimates, retries, wall-clock time, and success/failure.
  • Use Cases: Drive feature implementation by AI agents, orchestrate autonomous refactor loops, and run production-grade continuous agent workflows with quality gates.

Quick Start

Use the agentic-engineering skill to decompose a feature into 15-minute agent tasks, assign model tiers, and define capability and regression evals to verify delivery.

Frequently Asked Questions about agentic-engineering

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

FAQPage Schema
How do I execute feature implementation using AI agents safely?▼

Execute feature implementation using AI agents by applying an eval-first approach, decomposing work into 15-minute verifiable tasks, and routing by model complexity. This ensures quality, safety, and measurable outcomes for agent-led engineering loops.

What is eval-first execution for AI engineering work?▼

Eval-first execution is defining capability and regression evaluations before implementation to measure deltas and prevent regressions. This approach ensures AI agents produce verifiable units with clear done conditions and controlled costs.

How do I break down engineering work for autonomous agent loops?▼

Break down engineering work into independently verifiable 15-minute units with clear done conditions and single dominant risks. Task decomposition allows autonomous agent loops to execute refactors and feature delivery with quality gates.

How do I manage AI agent costs during code generation and refactoring?▼

Manage AI agent costs by routing tasks across model tiers based on complexity and tracking metrics for tokens, retries, wall-clock time, and success. Model routing rules ensure cost discipline during autonomous refactor loops and feature implementation.

Can I use this approach for production-grade continuous agent workflows?▼

Yes, this approach drives production-grade continuous agent workflows by applying capability eval definitions, 15-minute decomposed tasks, model-tier routing rules, and metrics tracking for tokens, retries, and success to ensure quality gates are met.

Why define regression evals before AI agents generate code?▼

Define regression evals before AI agents generate code to measure deltas and prevent regressions. Eval-first execution validates verifiable units against clear done conditions, ensuring autonomous agent loops maintain quality during refactors and feature delivery.

Related Skills

Explore More Software Engineering Skills→
shuyu-labsshuyu-labs
288

distributed-task-orchestrator

Decompose complex requests into atomic tasks and orchestrate parallel sub-agents.

Official
Advanced
sourcegraphsourcegraph
32

Evaluate Benchmark Traces

Evaluate AI agent benchmark traces for data integrity, output quality, and efficiency.

Official
Advanced
ryasrkryasrk

agentic-engineering

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

Community
Advanced
jfrometajfrometa

plan-writing

Decompose complex software work into small, verifiable tasks with dependencies.

Community
Basic