SDD Fidelity Review FAQs

Question 1

How does it categorize implementation deviations?

Accepted Answer

The skill uses multi-perspective AI analysis to categorize code as an exact match, a minor variation, a major difference, or missing functionality. Each deviation is assigned an impact level (low, medium, or high) to help developers prioritize fixes.

Question 2

Can this skill handle large project specifications?

Accepted Answer

Yes. It leverages a dedicated CLI tool to access specification data efficiently. This prevents context window bloat and allows it to process large JSON specs (50KB+) without wasting tokens or degrading performance.

Question 3

When should I use the SDD Fidelity Review skill?

Accepted Answer

This skill is best used during phase completion checkpoints, before merging pull requests, or after finishing complex tasks. It acts as an automated auditor to catch implementation drift before it compounds in later development phases.

Question 4

What is the SDD Fidelity Review skill?

Accepted Answer

The SDD Fidelity Review is a specialized Claude Code capability designed for Spec-Driven Development (SDD). It evaluates implementation code directly against specification requirements to ensure that the code built matches the original architectural plan.

Question 5

What kind of output does the fidelity review provide?

Accepted Answer

It generates structured JSON and Markdown reports. These reports include a fidelity score, detailed findings per task, documented vs. undocumented deviations, and specific recommendations for bringing the code back into alignment with the spec.

SDD Fidelity Review

SDD Fidelity Review

主要功能

使用场景

主要功能

使用场景