confidence-check_skill

Overview

This skill performs a pre-implementation confidence assessment and requires at least 90% confidence before starting any implementation. It prevents wasted work by checking for duplicates, verifying architecture alignment, confirming official documentation, finding working OSS references, and identifying root causes. Use it as a gate to ensure readiness and reduce rework. The skill returns a clear pass/warn/fail outcome and a detailed checklist.

How this skill works

The skill runs five weighted checks (duplicate detection, architecture compliance, official docs verification, OSS reference discovery, and root cause analysis) and computes a numeric confidence score. It searches the codebase for similar functionality, inspects planned architecture and stack requirements, fetches and verifies relevant official docs and APIs, locates proven open-source implementations, and analyzes logs/errors to confirm the underlying issue. If the composite score is >= 0.90 the skill marks readiness to proceed; lower scores produce alternatives or a stop recommendation.

When to use it

• Before beginning any new implementation or feature work
• When scope or requirements are unclear
• When a change could introduce new dependencies or architecture shifts
• When debugging recurring issues to confirm root cause
• Before approving developer work or issuing implementation tickets

Best practices

• Run the check as the first step in the implementation lifecycle to avoid wasted effort
• Use precise search terms and codegrep across the entire repository to detect duplicates
• Compare proposed changes against documented architecture and existing patterns to prevent unnecessary dependencies
• Always fetch and validate official API docs and changelogs rather than relying on memory
• Document OSS references and minimal reproducible examples used to justify proceeding

Example use cases

• A developer proposes a new module—confirm no equivalent exists and that it fits the tech stack before coding
• Triage a production bug—verify root cause and check for prior fixes or OSS patterns before implementing a hotfix
• Introduce third-party integration—validate official API compatibility and find sample implementations first
• Refactor request—ensure refactor aligns with architecture docs and does not duplicate behavior