GSD Verifier

Validates implemented code against phase requirements using empirical evidence checks.

Updated Mar 11, 2026
One-click install
npx skills add https://github.com/apsulli/gsd-extended --skill gsd-verifier-apsulli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: GSD Verifier
Source: https://github.com/apsulli/gsd-extended/tree/main/.agent/skills/verifier
Command: npx skills add https://github.com/apsulli/gsd-extended --skill gsd-verifier-apsulli

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI-assisted coding often produces work that claims to be complete but contains stubs, placeholders, or unwired components. This Skill verifies that implemented work actually achieves the stated phase goal through empirical evidence rather than trusting summary claims. ## Core Features & Use Cases - Three-Level Artifact Verification: Checks that required files exist, contain substantive implementations (not stubs), and are properly wired to other components. - Stub and Anti-Pattern Detection: Scans for TODO comments, placeholder text, empty implementations, and disconnected wiring patterns in React components and API routes. - Re-Verification Mode: Detects previous VERIFICATION.md files and runs optimized regression checks on passed items while fully re-verifying failed gaps. - Use Case: After an AI assistant completes a chat feature phase, run verification to confirm Chat.tsx actually fetches from the API, renders real message state, and contains no placeholder handlers before marking the phase done. ## Quick Start Ask the AI to verify phase 3 of the project against its must-haves and produce a VERIFICATION.md report.

Frequently Asked Questions about GSD Verifier

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

FAQPage Schema
How do I verify AI-generated code actually works?▼

Run the verification process which checks each required artifact at three levels: existence, substantive implementation, and proper wiring. It produces a VERIFICATION.md report with per-truth status, evidence, and structured gaps for any failures.

How to detect stub code and placeholder implementations?▼

The verifier scans for TODO/FIXME comments, placeholder text, empty return statements, and handlers that only call preventDefault or console.log. Findings are categorized as blockers, warnings, or info based on their impact on the phase goal.

What are must-haves in phase verification?▼

Must-haves are the truths (observable user behaviors), artifacts (concrete files), and key links (component wiring) required for a phase goal. They come from PLAN frontmatter or are derived from the phase goal in ROADMAP.md.

Does the verifier support re-verification after gap fixes?▼

Yes, when a previous VERIFICATION.md with gaps exists, it enters re-verification mode. Failed items receive full three-level verification while previously passed items get quick regression checks only.

What cannot be verified programmatically?▼

Visual appearance, user flow completion, real-time behavior like WebSockets, external service integration, and performance feel require human testing. The verifier outputs structured human verification items with test steps and expected results.