Why use a sub-agent for verification instead of running commands directly?

Verification commands often produce thousands of lines of logs. Using a sub-agent filters this output to just the essential errors (under 2k tokens), preventing your main Claude context from being flooded and losing focus.

What happens when a verification check fails?

The skill provides a concise summary of the failure, including the first few lines of the error log. This allows the main agent to analyze the problem and implement a fix without reading the entire verbose output.

What languages and frameworks are supported?

It is language-agnostic. It can execute any command available in your environment, including npm/yarn, pytest, go test, make, and custom shell scripts.

How does the skill know which test commands to run?

The skill references a 'commands.md' file, typically created by the 'discover-project-commands' skill, which catalogs project-specific scripts for testing, linting, and building.

Verify Implementation

Name: Verify Implementation
Author: eveld

byeveld

•

セキュリティとテスト

Automates post-implementation validation by running tests, linting, and build commands through a context-efficient sub-agent.

The Verify Implementation skill provides a robust workflow for validating code changes within the Claude Code environment. By spawning a specialized Bash sub-agent, it executes automated checks—including unit tests, linting, type checking, and build scripts—while shielding the main conversation context from voluminous log output. This approach ensures that only essential pass/fail statuses and concise error summaries are returned to the main agent, maintaining high token efficiency and preventing the model from losing focus during complex development tasks.

主な機能

01Sub-agent orchestration for token-efficient verification

02Structured pass/fail reporting for better decision making

03Project-specific command discovery and execution

043 GitHub stars

05Intelligent error log summarization and filtering

06Automated execution of tests, linting, and builds

ユースケース

01Running comprehensive linting and type checks on newly generated features

02Validating bug fixes against a test suite before finalizing changes

03Ensuring build stability in complex monorepos without flooding context

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add eveld/claude verify-implementation

For use in Claude.ai and ChatGPT