How do I trigger a document audit?

You can trigger the audit by using keywords like 'SOW review', 'design review', or '90 points scoring' within your Claude Code environment.

What are the four scoring dimensions evaluated?

The skill evaluates Accuracy (fact vs. inference), Completeness (coverage of requirements), Relevance (alignment with goals), and Actionability (implementation clarity).

How do the reliability markers (✓/→/?) function?

These markers are used to indicate the level of evidence: ✓ for verified facts, → for items requiring correction or improvement, and ? for missing or unknown information.

Can this skill check consistency between different files?

Yes, it includes a specific SOW ↔ Spec consistency check to ensure that acceptance criteria in the SOW map correctly to functional requirements and test plans in the Spec.

What is the pass threshold for a document review?

A document must score at least 90 out of 100 to pass the quality gate. Scores below 90 require revisions and a re-review before proceeding to the implementation phase.

Design Document Audit Scoring

Name: Design Document Audit Scoring
Author: thkt

bythkt

•

セキュリティとテスト

Evaluates technical design documents and Statements of Work (SOW) using a rigorous 100-point scoring system to ensure high-quality implementation specs.

This skill provides a standardized framework for reviewing technical design documents (SOW/Spec) by scoring them across four critical dimensions: accuracy, completeness, relevance, and actionability. By enforcing a 90-point quality gate and utilizing reliability markers (✓/→/?), it helps developers and project managers identify gaps in specifications before implementation begins. This objective audit process reduces the risk of costly rework and misaligned expectations by ensuring that every requirement is backed by evidence and actionable implementation steps.

主な機能

01Strict 90-point quality gate threshold for implementation readiness

02Automated consistency checks between SOW and Spec documentation

033 GitHub stars

04100-point scoring system across four key quality dimensions

05Reliability markers (✓/→/?) for transparent evidence-based feedback

06Standardized templates for professional review reporting and action items

ユースケース

01Objective scoring of technical design proposals to identify risks

02Bridging the gap between project requirements and actionable implementation steps

03Pre-implementation quality checks for SOW and technical specification documents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add thkt/claude-config formatting-audits

For use in Claude.ai and ChatGPT

Download Skill