skills.rest
cescrafli
cescrafliCOMMUNITY
Explore all skills by this author →
GitHub

agentic-engineering

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

Updated Mar 26, 2026
Search Tags:#decomposition#risk-management#agentic#ai-workflows#model-routing#eval-first#cost-discipline
One-click install
npx skills add https://github.com/cescrafli/compyrasion --skill agentic-engineering-cescrafli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentic-engineering
Source: https://github.com/cescrafli/compyrasion/tree/main/skills/agentic-engineering
Command: npx skills add https://github.com/cescrafli/compyrasion --skill agentic-engineering-cescrafli

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Operate as an agentic engineer using eval-first execution, decomposition, and cost-aware model routing.

Core Features & Use Cases

  • Eval-first loop: define capability eval and regression eval, run baseline, execute implementation, re-run evals, and compare deltas.
  • Task decomposition: apply the 15-minute unit rule; each unit independently verifiable, with a single dominant risk, and a clear done condition.
  • Model routing and session strategy: route by task complexity (Haiku, Sonnet, Opus) and manage sessions through transitions to minimize friction.
  • Review focus and cost discipline: emphasize invariants, edge cases, security, and track per-task model, tokens, retries, and time.
  • Cost discipline: track model usage and costs; escalate tiers only when needed.

Quick Start

Plan and execute AI-driven engineering tasks by decomposing work into verifiable units and selecting model tiers based on 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 costs when routing AI engineering tasks across different model tiers?▼

Manage AI engineering costs by routing tasks based on complexity, escalating model tiers only when needed, and tracking per-task tokens, retries, and time to maintain cost discipline.

What is an eval-first workflow for agentic engineering?▼

An eval-first workflow defines capability and regression evals, runs a baseline, executes implementation, and re-runs evals to compare deltas, ensuring predictable delivery for AI-driven tasks.

How do I decompose multi-file engineering tasks for AI execution?▼

Decompose multi-file engineering tasks using the 15-minute unit rule, ensuring each unit is independently verifiable, has a single dominant risk, and a clear done condition.

Can I use this approach for iterative development and risk-managed deployment?▼

Yes, this approach coordinates iterative development and risk-managed deployment by requiring upfront completion criteria and verifiable task units to govern cross-file collaboration.

What's the best way to handle session transitions in AI workflows to minimize friction?▼

Handle AI workflow session transitions by managing model routing strategies across complexity tiers, minimizing friction while maintaining execution continuity across verifiable task units.

Why does my AI workflow execution lack predictable delivery?▼

AI workflow execution lacks predictable delivery without upfront completion criteria, verifiable task units, and cost-aware model tier routing to govern the implementation and review process.

Related Skills

Explore More Software Engineering Skills→
xu-xiangxu-xiang
1.8k

eval-harness

Define, execute, and report capability and regression evals for Claude Code sessions.

Community
Intermediate
T-ravT-rav
5

eval-harness

Automate Claude Code session evaluations with capability and regression tests.

Community
Advanced
zhmxiaowozhmxiaowo
2

agentic-engineering

Coordinate AI agents to implement engineering tasks with eval-first validation.

Community
Advanced
jfrometajfrometa

plan-writing

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

Community
Basic