Does it support different programming languages?

Yes, the skill uses universal search and read commands that work across any language or framework within the Claude Code environment.

How does this skill prevent inaccurate documentation?

It enforces a mandatory verification workflow that includes searching for patterns, reading file contents, and checking for alternative implementations before allowing a 'missing' or 'implemented' claim.

Is this skill part of a larger framework?

Yes, it is part of the Lore plugin framework, which provides structured and deterministic workflows for AI-assisted development.

Can this skill help with legacy code audits?

Yes, it is specifically designed for situations where documentation and code may have diverged, allowing the AI to verify the ground truth of the current implementation.

What is the Verify Before Documenting skill?

It is a specialized Claude Code capability designed to prevent AI hallucinations by requiring the model to search, read, and cite actual source code before making assertions about its state.

Evidence-Based Code Documentation

Name: Evidence-Based Code Documentation
Author: LayerDynamics

byLayerDynamics

0•

Learning & Documentation

Enforces rigorous file verification and evidence gathering to prevent false claims and inaccurate documentation in codebase audits.

The Verify Before Documenting skill transforms Claude into a high-precision codebase auditor by enforcing an evidence-based workflow. It prevents the common AI pitfall of making assumptions about code existence or functionality by requiring systematic searches and direct file reads before any claims are documented. This skill is essential for conducting technical gap analyses, comparing legacy documentation to current implementations, and performing thorough code reviews where accuracy is paramount. By citing specific file paths and line numbers, it ensures that every finding is grounded in the actual source code rather than hallucinated patterns.

Key Features

010 GitHub stars

02Structured 6-step verification workflow from identification to documented evidence

03Alternative implementation detection to distinguish between missing and renamed features

04Integrated search patterns for dependencies, imports, and code structures

05Standardized documentation templates for reporting verified implementation status

06Evidence-based citation requirements for file paths and specific code excerpts

Use Cases

01Auditing legacy codebases to synchronize outdated documentation with the actual source state

02Performing a technical gap analysis to identify missing features versus alternative implementations

03Verifying implementation requirements before finalizing status reports or pull request descriptions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add layerdynamics/lore verify-before-documenting

For use in Claude.ai and ChatGPT

Key Features

010 GitHub stars

02Structured 6-step verification workflow from identification to documented evidence

03Alternative implementation detection to distinguish between missing and renamed features

04Integrated search patterns for dependencies, imports, and code structures

05Standardized documentation templates for reporting verified implementation status

06Evidence-based citation requirements for file paths and specific code excerpts

Use Cases

01Auditing legacy codebases to synchronize outdated documentation with the actual source state

02Performing a technical gap analysis to identify missing features versus alternative implementations

03Verifying implementation requirements before finalizing status reports or pull request descriptions