How does this skill help save on token costs?

By investing 100-200 tokens in a confidence check, you can avoid spending 5,000-50,000 tokens on incorrect, redundant, or hallucinated code implementations.

What criteria are used to determine the confidence score?

It evaluates five areas: no duplicate implementation (25%), architecture compliance (25%), official documentation verification (20%), OSS reference (15%), and root cause identification (15%).

What happens if the confidence score is below 70%?

The framework recommends stopping implementation immediately and requesting additional context or clarification to prevent errors.

What is the primary purpose of the Confidence Check skill?

The primary purpose is to evaluate the reliability of a proposed implementation before code is written, ensuring the AI has enough context and the correct direction.

Confidence Check

Name: Confidence Check
Author: wishdevs

bywishdevs

0•

생산성 및 워크플로

Evaluates task implementation reliability through a weighted scoring framework to prevent incorrect code paths and optimize token usage.

Confidence Check is a strategic assessment framework for Claude Code designed to ensure high-precision development by validating implementation plans before coding begins. By scoring tasks against five critical criteria—including architectural compliance, duplicate detection, and root cause identification—it enforces a strict 90% confidence threshold. This proactive approach prevents wasted effort on redundant or misaligned features, offering an impressive ROI by spending a few hundred tokens to save tens of thousands in potential rework.

주요 기능

01Weighted 5-point evaluation system for objective reliability scoring

02Architecture alignment verification for specific tech stacks like Supabase and pytest

03Token-efficient decision framework with clear Go/Clarify/Stop recommendations

04Automated duplicate implementation detection using grep and glob tools

05Integration with MCP tools for real-time documentation and OSS reference checking

060 GitHub stars

사용 사례

01Checking if a new feature request overlaps with existing functionality to avoid technical debt

02Validating a bug fix approach before modifying the codebase to ensure the root cause is addressed

03Ensuring complex refactoring tasks align with established project architecture and documentation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add wishdevs/wishdev-marketplace confidence-check

For use in Claude.ai and ChatGPT

Download Skill