How does this skill handle subagent delegation?

It instructs the agent to never trust a subagent's 'success' report at face value. Instead, it must independently check the VCS diff and run verification commands on the changes.

What happens if a verification command fails?

The agent is required to report the actual status with evidence of the failure rather than claiming success, allowing for immediate course correction.

Can I bypass verification if the change is small?

No. The skill is built on the philosophy of 'no exceptions,' treating partial or skipped verification as an unverified and therefore invalid claim.

Why is this skill necessary for Claude Code?

It prevents the AI from assuming work is finished without proof, ensuring that all claims of success are backed by actual command output rather than 'hallucinated' confidence.

What counts as a valid verification?

Valid verification requires fresh output from tests, linters, or build commands showing zero failures and a successful exit code (exit 0) specifically for the current state of the code.

Verification Before Completion

Name: Verification Before Completion
Author: LayerDynamics

byLayerDynamics

0•

セキュリティとテスト

Enforces a strict discipline of running fresh verification commands and analyzing actual output before claiming any task is finished or successful.

The Verification Before Completion skill establishes a non-negotiable standard for AI-assisted development, preventing 'hallucinated' success reports. By requiring evidence through fresh command execution, it ensures that tests, linters, builds, and requirement checks have actually passed before an agent claims completion. This skill transforms the workflow from assumption-based to evidence-based, significantly reducing regression rates and improving the reliability of subagent delegations by mandating an independent check of all outputs and diffs.

主な機能

01Mandatory fresh verification command execution

02Comprehensive checklist for diverse claim types (Tests, Build, Linter)

03Independent verification of subagent success reports

040 GitHub stars

05Structured output analysis for exit codes and failure counts

06Verification-gated completion claims

ユースケース

01Validating that bug fixes actually resolve the original failing scenario

02Confirming build stability and linting compliance in CI/CD pre-checks

03Ensuring all unit and integration tests pass before a final submission

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add layerdynamics/lore verification-before-completion

For use in Claude.ai and ChatGPT

主な機能

01Mandatory fresh verification command execution

02Comprehensive checklist for diverse claim types (Tests, Build, Linter)

03Independent verification of subagent success reports

040 GitHub stars

05Structured output analysis for exit codes and failure counts

06Verification-gated completion claims

ユースケース

01Validating that bug fixes actually resolve the original failing scenario

02Confirming build stability and linting compliance in CI/CD pre-checks

03Ensuring all unit and integration tests pass before a final submission

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add layerdynamics/lore verification-before-completion

For use in Claude.ai and ChatGPT