skills.rest
samymity
samymityCOMMUNITY
Explore all skills by this author →
GitHub

agentic-engineering

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

1|Updated Mar 3, 2026
Search Tags:#agentic engineering#regression testing#model routing#eval-first#code decomposition#cost discipline#quality controls
One-click install
npx skills add https://github.com/samymity/bridge-ventures-backend --skill agentic-engineering-samymity
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentic-engineering
Source: https://github.com/samymity/bridge-ventures-backend/tree/main/.claude/skills/agentic-engineering
Command: npx skills add https://github.com/samymity/bridge-ventures-backend --skill agentic-engineering-samymity

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps teams reduce the risk of broken implementations by enforcing eval-first execution, careful decomposition, and cost-aware model routing for agent-driven development.

Core Features & Use Cases

  • Eval-First Loop: Run baseline and regression evals, implement changes, then re-run checks to confirm deltas and catch regressions.
  • Agent-Sized Decomposition: Break work into independently verifiable units with clear done conditions and dominant single risks.
  • Model Routing by Complexity: Match task difficulty to Haiku, Sonnet, or Opus to balance quality and cost.
  • Review & Cost Discipline: Focus human review on invariants, edge cases, security/auth assumptions, and rollout risk while tracking model usage, retries, and wall-clock time.

Quick Start

Ask your AI agent to apply eval-first execution and decompose the feature into independently verifiable units, routing tasks to the right model tier while tracking cost and validating with regression evals.

Frequently Asked Questions about agentic-engineering

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

FAQPage Schema
How do I prevent regressions when using AI agents for multi-file code refactoring?▼

Eval-first agent engineering prevents regressions by running baseline evals before changes and regression evals after, verifying deltas to catch issues in multi-file refactoring workflows.

What is eval-first development and how does it work with AI coding agents?▼

Eval-first development runs baseline evals, implements changes, then re-runs regression evals to confirm deltas. This loop catches AI agent regressions immediately during software implementation.

How do I balance model costs and quality when routing tasks to AI coding agents?▼

Model routing by complexity balances costs and quality by matching task difficulty to appropriate model tiers like Haiku, Sonnet, or Opus, enforcing cost-aware escalation rules.

How do I decompose large software features into safe units for AI agent implementation?▼

Agent-sized decomposition breaks features into independently verifiable units with clear done conditions and dominant single risks, enforcing 15-minute unit granularity for safe AI implementation.

What should human reviewers focus on when checking AI-generated code changes?▼

Human review for AI-generated code should focus on invariants, edge cases, security and auth assumptions, and rollout risk, while tracking model usage, retries, and wall-clock time.

When should I avoid using autonomous agents for software implementation?▼

Avoid autonomous agents for tasks lacking clear completion criteria or baseline evals, as eval-first execution requires defined done conditions and measurable regression checks to ensure safety.

Related Skills

Explore More Software Engineering Skills→
rongarederongarede
4

eval-harness

Evaluate Claude Code sessions with capability and regression evals.

Community
Advanced
dzackgarzadzackgarza
1

difficulty-and-time-estimation

Estimate task complexity using token budgets, turn count, and risk metrics.

Community
Intermediate
robit-manrobit-man

regression-cicd-hooks

Automate regression testing in CI/CD pipelines with baseline comparison and merge blocking.

Community
Advanced
ajstack22ajstack22

atlas-meta

Analyze task complexity and route software development tasks to Atlas workflow tiers.

Community
Basic