spec-walkthrough

Briefs founders on completed specs with result changes and manual test plans.

1|1|Updated May 11, 2026
One-click install
npx skills add https://github.com/taxmaxi/taxmaxi --skill spec-walkthrough-taxmaxi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: spec-walkthrough
Source: https://github.com/taxmaxi/taxmaxi/tree/main/.agents/skills/spec-walkthrough
Command: npx skills add https://github.com/taxmaxi/taxmaxi --skill spec-walkthrough-taxmaxi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Founders who did not watch a spec's implementation cannot tell what the product now does differently, which changed numbers are intentional, or what to test by hand. This Skill turns a finished or in-flight spec into a tech-lead briefing that transfers exactly that knowledge. ## Core Features & Use Cases - State-based reconstruction: Reads the issue body, merged PRs, report comments, and ADRs instead of reconstructing the story from the comment event log. - PR triage into three bins: Classifies merged work as result-changing, new surface, or result-preserving so testing effort goes only where behavior actually changed. - Ranked manual test plan: Derives one concrete check per result change and new surface, with expected outcomes, expected blockers, and risk-ranked ordering. - Use Case: After a tax-calculation spec ships, ask for a walkthrough and receive the story, the numbers that changed on purpose with links to their recorded decisions, a checklist of manual tests against real local data, known gaps already filed, and honest confessions about untested paths. ## Quick Start Ask the agent to walk you through what spec number 42 delivered and what you should manually test.

Frequently Asked Questions about spec-walkthrough

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I catch up on what a finished spec delivered?▼

Ask for a spec walkthrough. The Skill reads the issue body, merged PRs, and report comments, then briefs you on what the product does differently, which numbers changed on purpose, and what to test by hand.

How to build a manual test plan from merged pull requests?▼

The Skill sorts merged PRs into result-changing, new surface, and result-preserving bins, then derives one concrete check per result change and new surface, ranked with the riskiest checks first and expected blockers noted.

Does the spec walkthrough work for in-flight specs?▼

Yes. For in-flight specs it distinguishes proposed work from merged and deployed behavior, and states plainly whether the work is live in production or requires a local dev run to test.

Does the walkthrough change code, issues, or project state?▼

No. The walkthrough is strictly a briefing. It reads issues, PRs, ADRs, and deployment state but modifies nothing, and it can optionally save the manual test plan as a local file.

Why does the walkthrough skip some merged PRs in the test plan?▼

Result-preserving PRs, such as refactors proven identical by differential or parity tests, get one line and no manual testing. Only result-changing and new-surface work generates test checks.