the-craft

Define an evidence-based operating posture for AI-assisted codebase work.

Updated Jun 10, 2026
One-click install
npx skills add https://github.com/brewpirate/acme-frontier-ai --skill the-craft
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: the-craft
Source: https://github.com/brewpirate/acme-frontier-ai/tree/main/catalog/projects/the-craft/skills/the-craft
Command: npx skills add https://github.com/brewpirate/acme-frontier-ai --skill the-craft

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill gives you a clear operating posture for substantial coding work with an AI collaborator, reducing guesswork, skipped verification, and sloppy handoffs.

Core Features & Use Cases

  • Evidence-first workflow: Treats tool output as proof and requires verification before claiming success.
  • Collaboration guardrails: Sets direct interaction rules for working with the user, surfacing uncertainty, and asking before continuing.
  • Conflict and quality handling: Explains how to resolve rule conflicts, stop on surprises, and fix broken windows without expanding scope.
  • Use Case: Use it when starting a large implementation, reviewing a tricky change, or coordinating a multi-step coding session where correctness matters more than speed.

Quick Start

Use the craft skill to guide how we should approach this codebase task from start to finish.

Frequently Asked Questions about the-craft

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

FAQPage Schema
How do I enforce evidence-based verification before declaring an AI coding task complete?

Evidence-based verification requires treating tool output as proof and validating changes before claiming success. This Skill establishes an operating posture that enforces verification, conflict resolution, and broken-window handling to prevent sloppy handoffs in substantial software tasks.

What is the best way to manage rule conflicts during AI-assisted codebase collaboration?

Managing rule conflicts requires task-specific rule selection and direct interaction guardrails. This Skill provides collaboration guardrails that surface uncertainty, resolve conflicts systematically, and stop on surprises without expanding the original task scope.

How do I set stop conditions for multi-step software engineering tasks with an AI agent?

Setting stop conditions involves defining quality checks and surprise-handling protocols before work begins. This Skill defines explicit stop conditions and broken-window handling rules to ensure correctness is prioritized over speed during large implementation sessions.

When do I need an operating posture for substantial coding work with an AI collaborator?

An operating posture is needed when starting large implementations, reviewing tricky changes, or coordinating multi-step coding sessions. This Skill applies to substantial repository work where correctness matters and requires evidence-based verification before completion.

How do I fix broken windows in a codebase without expanding the current task scope?

Fixing broken windows requires stopping to address issues immediately without scope creep. This Skill explains how to handle broken windows by resolving the defect directly and maintaining task-specific rule selection to keep the original work focused.

Can I use this workflow for quick code edits instead of substantial software tasks?

This workflow is designed for substantial software tasks, collaboration sessions, and tricky changes where correctness matters more than speed. Quick edits do not require the full operating posture of evidence-based verification and rule conflict resolution defined by this Skill.