wow-loop

Orchestrates a multi-agent implement-verify-fix loop with screenshot evidence gates for deliverables.

Updated Aug 30, 2026
One-click install
npx skills add https://github.com/ryanportfolio/threejs-interview-test --skill wow-loop-ryanportfolio
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: wow-loop
Source: https://github.com/ryanportfolio/threejs-interview-test/tree/main/.claude/skills/wow-loop
Command: npx skills add https://github.com/ryanportfolio/threejs-interview-test --skill wow-loop-ryanportfolio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Deliverables often ship because they look finished rather than because they survived rigorous review. This Skill replaces subjective sign-off with an adversarial, evidence-gated loop where verifiers armed with screenshots and measurements must fail to break the work before it ships. ## Core Features & Use Cases - Structured multi-agent pipeline: Recon, single-director spec, sole implementer, adversarial verification, fix rounds, and a final orchestrator gate. - Deterministic capture hooks: Freeze the deliverable at named states so screenshots are byte-identical and verdicts are reproducible. - Checkable quality bar: Distills a reference example into 5-7 measurable mechanisms in bar.md so reviewers verify facts instead of vibes. - Use Case: Polishing a WebGL intro, landing page, slide deck, or PDF until independent verifiers with fresh captures find zero confirmed defects. ## Quick Start Invoke the wow-loop skill with the target deliverable, for example by typing /wow-loop followed by the name of the page or document you want perfected.

Frequently Asked Questions about wow-loop

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

FAQPage Schema
How do I run an adversarial review loop on a landing page?▼

Invoke the loop with the page as its target. It runs recon, writes one spec with a checkable bar, has a single implementer build it, then sends experience and engineering verifiers with fresh screenshots to disprove the work across fix rounds.

What is evidence-gated verification for AI-generated deliverables?▼

Evidence-gated verification means work ships only when verifiers armed with screenshots, measurements, and test runs fail to find confirmed defects. Claims are untrusted until backed by named evidence paths, and verdicts are binary pass or fail per finding.

Can this loop review PDFs and slide decks, not just websites?▼

Yes, the loop is subject-agnostic and works for WebGL intros, landing pages, slide decks, and PDFs. Before rounds begin it preflights a capture path for that deliverable type, such as PDF page renders or video filmstrips.

Why does the loop require deterministic screenshots?▼

Without reproducible frames, verifier verdicts are guesses. The loop mandates seeded randomness and a capture hook that freezes the deliverable at named states so a screenshot taken twice is byte-identical.

When does the fix-and-verify loop stop?▼

It stops when a round produces zero confirmed findings or only honestly named residuals with causes, capped at about five rounds. It never ends merely because the work looks done.