How does this skill save tokens?

By spending a few hundred tokens on a pre-check, you avoid generating thousands of tokens of incorrect or redundant code that would later need to be refactored or deleted.

Why do I need a 90% confidence score?

A high threshold ensures that the AI has fully understood the context and verified all dependencies, preventing 'wrong-direction' work that wastes time and tokens.

Does it search my local codebase?

Yes, it uses tools like Grep and Glob to scan your existing files for duplicate functionality and to ensure alignment with your current architectural patterns.

What happens if the confidence score is below 70%?

The skill will trigger a stop command, requesting more context or documentation from the user rather than attempting an implementation that is likely to fail.

Confidence Check

Name: Confidence Check
Author: Microck

byMicrock

•

部署与 DevOps

Evaluates implementation readiness through a rigorous 90% confidence threshold to prevent redundant or architecturally misaligned development.

关于

The Confidence Check skill acts as a critical guardrail for AI-driven development by assessing five key technical criteria before any code is written. By verifying architecture compliance, searching for existing duplicate implementations, and cross-referencing official documentation and open-source patterns, it ensures that Claude proceeds only when there is a 90% or higher probability of success. This preemptive assessment drastically reduces hallucination risks and token wastage by identifying root causes and potential blockers before they become costly implementation errors.

主要功能

Calculates a weighted confidence score (0.0 - 1.0) based on five critical success factors
Identifies underlying root causes to ensure developers solve problems, not just symptoms
Cross-references implementation plans with official documentation and proven open-source examples
81 GitHub stars
Performs automated duplicate checks to prevent redundant code across the repository
Verifies alignment with project architecture, CLAUDE.md guidelines, and existing tech stacks

使用场景

Before starting a major refactor or new feature implementation to ensure architectural consistency
During bug fixing to confirm the root cause is understood before applying a patch
When exploring new libraries to verify compatibility and reference proven community patterns

关于

主要功能

Calculates a weighted confidence score (0.0 - 1.0) based on five critical success factors
Identifies underlying root causes to ensure developers solve problems, not just symptoms
Cross-references implementation plans with official documentation and proven open-source examples
81 GitHub stars
Performs automated duplicate checks to prevent redundant code across the repository
Verifies alignment with project architecture, CLAUDE.md guidelines, and existing tech stacks

使用场景

Before starting a major refactor or new feature implementation to ensure architectural consistency
During bug fixing to confirm the root cause is understood before applying a patch
When exploring new libraries to verify compatibility and reference proven community patterns