agentic-engineering

Coordinate agentic engineering workflows with eval-first execution, decomposition, and tiered model routing.

1|Updated Mar 18, 2026
One-click install
npx skills add https://github.com/ValenBorras/upscribe --skill agentic-engineering-valenborras
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agentic-engineering
Source: https://github.com/ValenBorras/upscribe/tree/main/.github/skills/agentic-engineering
Command: npx skills add https://github.com/ValenBorras/upscribe --skill agentic-engineering-valenborras

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides a structured framework for AI-driven software tasks, enabling eval-first execution, disciplined decomposition, and cost-aware model routing to improve reliability, safety, and efficiency.

Core Features & Use Cases

  • Eval-First Loop: define capability evals and regression checks, run baselines, implement changes, and re-evaluate deltas.
  • Task Decomposition & Session Strategy: break work into independently verifiable units, manage tightly scoped risk, and maintain sessions across milestones.
  • Model Routing & Review: route tasks by complexity across Haiku, Sonnet, and Opus tiers, with rigorous review focusing on invariants, edge cases, and security considerations.

Quick Start

Decompose a given engineering task, select model tiers by complexity, and run evals before implementing changes.

Frequently Asked Questions about agentic-engineering

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

FAQPage Schema
What is eval-first routing in AI-driven software engineering?

AI model routing assigns engineering tasks to Haiku, Sonnet, or Opus tiers based on complexity, balancing cost and capability. This tiered routing ensures simple tasks use efficient models while complex tasks get advanced processing.

How do I decompose software engineering tasks for multiple AI sessions?

Use task decomposition and session strategy to divide work into independently verifiable units. Manage tightly scoped risk and maintain sessions across milestones for incremental implementation.

How do I track token and model costs for agentic engineering workflows?

Implement cost-tracking for models, tokens, retries, and runtime. Cost-aware model routing across Haiku, Sonnet, and Opus tiers manages expenses while maintaining rigorous quality checks.

Can I use tiered model routing for complex architecture planning?

Yes, route architecture planning tasks by complexity across Haiku, Sonnet, and Opus tiers. The framework applies rigorous review focusing on invariants, edge cases, and security considerations.

What limitations exist with eval-first execution for AI workflows?

Eval-first execution requires explicit completion criteria and unit-decomposition rules to function. Without these prerequisites and guardrails for review and risk management, the structured framework cannot ensure reliability.