What is the minimum confidence score required to proceed?

The skill requires a minimum score of 90% (0.90) to proceed with implementation, ensuring high reliability and architectural alignment.

How does this skill help save on token costs?

It spends a small amount (100-200 tokens) on a pre-implementation check to prevent expensive (5,000-50,000 tokens) errors and 'wrong-direction' code generation.

What happens if the confidence score is too low?

If the score is between 70% and 90%, it presents alternatives or asks clarifying questions; if it falls below 70%, it stops the task and requests more context.

Does it check for existing code within my repository?

Yes, the skill uses Grep and Glob tools to search your codebase for similar functions or modules to prevent redundant or duplicate implementations.

How does it verify technical requirements?

It utilizes WebSearch and WebFetch tools to reference official documentation and proven open-source implementations to ensure the proposed solution is valid.

Confidence Check

Name: Confidence Check
Author: SuperClaude-Org

bySuperClaude-Org

•

20,400

•

Productivity & Workflow

Validates implementation readiness through a multi-point scoring framework to prevent architectural drift and redundant code.

The Confidence Check skill acts as a high-precision guardrail for AI-driven development by enforcing a mandatory 90% confidence threshold before any code implementation begins. By systematically verifying architectural alignment, searching for existing codebase patterns, and validating against official documentation, it ensures that every proposed change is intentional, correctly architected, and necessary. This proactive assessment significantly reduces token waste and prevents the introduction of technical debt by stopping 'wrong-direction' work before it starts.

Key Features

01Weighted 5-point assessment scoring (0.0 - 1.0) with a 90% success threshold

02Root cause identification to ensure symptoms aren't confused with underlying issues

03Architecture compliance verification against CLAUDE.md and existing tech stacks

0420,400 GitHub stars

05Real-time official documentation validation via WebSearch and WebFetch

06Automated codebase search using Grep and Glob to prevent duplicate implementations

Use Cases

01Verifying API implementation details against current official documentation before coding

02Preventing duplicate utility functions in large, unfamiliar codebases

03Ensuring new features align with established architectural patterns and dependencies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add superclaude-org/superclaude_framework confidence-check

For use in Claude.ai and ChatGPT

Download Skill