How does the Evidence Verification skill prevent AI hallucinations?

It requires the AI to run actual shell commands—such as tests, builds, and linters—and capture the real output and exit codes rather than simply claiming a task is finished based on the code written.

What counts as 'evidence' in this skill?

Evidence includes command exit codes (where 0 indicates success), test pass/fail/skipped counts, code coverage percentages, build artifacts, linter results, and security scan outputs.

Can this skill be used for production deployment verification?

Yes, it includes specific protocols for capturing deployment status, verifying environment health check results, and documenting rollback capabilities.

Does it support multiple programming languages?

Absolutely. It provides standardized workflows and common commands for various ecosystems including JavaScript, TypeScript, Python, and more.

When should this skill be activated?

It auto-activates during code implementation completion, code reviews, and before agent handoffs, but can also be manually triggered when a user requests verification.

Evidence Verification

Name: Evidence Verification
Author: yonatangross

byyonatangross

•

セキュリティとテスト

Verifies development tasks and code changes by collecting verifiable proof including test results, build outputs, and exit codes.

概要

The Evidence Verification skill equips Claude with a rigorous protocol for validating task completion through tangible evidence. It ensures that every code change, deployment, or review is backed by executable proof—such as passing test suites, successful build logs, coverage metrics, and zero-exit codes—moving beyond verbal claims to production-grade verification. This skill is essential for maintaining high code quality, preventing AI hallucinations during the development cycle, and ensuring that implementations meet project standards before they are marked as complete.

主な機能

Automated collection of test results, exit codes, and coverage metrics
Multi-level verification standards from Minimum to Gold Standard
Standardized evidence templates for builds, tests, and code quality
Integration with language-specific toolchains like npm, pytest, and ruff
Protocol for capturing deployment health checks and security scans
29 GitHub stars

ユースケース

Automating quality assurance checks and linter verification during code reviews
Verifying that a new feature passes all unit and integration tests before handoff
Ensuring a build is successful and produces required artifacts before deployment

概要

主な機能

Automated collection of test results, exit codes, and coverage metrics
Multi-level verification standards from Minimum to Gold Standard
Standardized evidence templates for builds, tests, and code quality
Integration with language-specific toolchains like npm, pytest, and ruff
Protocol for capturing deployment health checks and security scans
29 GitHub stars

ユースケース

Automating quality assurance checks and linter verification during code reviews
Verifying that a new feature passes all unit and integration tests before handoff
Ensuring a build is successful and produces required artifacts before deployment