How does this improve the code review process?

It shifts the human reviewer's focus toward high-risk areas like invariants, security, and edge cases, leaving style enforcement to automated tools.

What is the 'Eval-First Loop' in this skill?

It is a methodology where capability and regression evals are defined and baselined before any implementation code is written to ensure measurable progress.

Does it include cost management features?

Yes, it incorporates cost discipline protocols to track model usage, token estimates, retries, and wall-clock time for every task.

How does this skill handle task decomposition?

It applies a '15-minute unit rule,' breaking work into small, independently verifiable tasks with a single dominant risk and a clear done condition.

Which models are used for different engineering tasks?

The skill routes simple edits to Haiku, standard implementation to Sonnet, and complex architecture or root-cause analysis to Opus to optimize cost and performance.

Agentic Engineering

Name: Agentic Engineering
Author: manoranjan30

bymanoranjan30

0•

Productividad y Flujo de Trabajo

Implements a structured, eval-first workflow for autonomous software engineering with intelligent task decomposition and model routing.

The Agentic Engineering skill transforms Claude into a disciplined AI agent by enforcing rigorous operating principles such as eval-first loops, 15-minute task decomposition, and cost-aware model routing. It optimizes the engineering lifecycle by prioritizing verification before execution, ensuring that AI agents handle heavy implementation work while humans maintain control over high-level risk and quality. By focusing on invariants, edge cases, and regression checks, this skill minimizes development drift and maximizes the efficiency and reliability of AI-driven coding sessions.

Características Principales

01Eval-first loop execution for verified implementations

02Structured session management and milestone compaction

03Systematic 15-minute unit task decomposition

04Cost-aware model routing across Haiku, Sonnet, and Opus

050 GitHub stars

06Risk-focused code review protocols for AI-generated code

Casos de Uso

01Building reliable features with regression-first testing loops

02Managing complex refactors using autonomous AI-driven workflows

03Scaling development velocity while maintaining strict architectural invariants

Características Principales

01Eval-first loop execution for verified implementations

02Structured session management and milestone compaction

03Systematic 15-minute unit task decomposition

04Cost-aware model routing across Haiku, Sonnet, and Opus

050 GitHub stars

06Risk-focused code review protocols for AI-generated code

Casos de Uso

01Building reliable features with regression-first testing loops

02Managing complex refactors using autonomous AI-driven workflows

03Scaling development velocity while maintaining strict architectural invariants