yak-sniff-test

Review completed work against the original brief for a binary verdict.

3|3|Updated Feb 13, 2026
One-click install
npx skills add https://github.com/mrdavidlaing/yakthang --skill yak-sniff-test
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: yak-sniff-test
Source: https://github.com/mrdavidlaing/yakthang/tree/main/skills/yak-sniff-test
Command: npx skills add https://github.com/mrdavidlaing/yakthang --skill yak-sniff-test

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Does this yak smell right? Use when a shaver reports done and Yakob needs to verify the work matches the brief using a fresh, independent reviewer agent before accepting or pruning the yak.

Core Features & Use Cases

  • Adversarial reviewer: spawns a fresh agent to review the done work against the brief.
  • Bias reduction: prevents anchoring by not letting the shaver's narrative influence the reviewer.
  • Verification workflow: provides a binary verdict to drive wip-state transitions (pass/needs-info/fail).

Quick Start

Invoke after a shaver reports completion to verify the work aligns with the brief.

Frequently Asked Questions about yak-sniff-test

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

FAQPage Schema
How do I perform an independent verification of completed work against the original brief?

Perform an independent verification by spawning a fresh reviewer agent to adversarially check completed work against the brief. This reduces bias by preventing the original worker's narrative from influencing the reviewer.

What is an adversarial review workflow for quality assurance?

An adversarial review workflow uses a fresh agent to validate completed work against the original brief, reducing anchoring bias. It produces a binary verdict to drive state transitions, ensuring work passes independent verification before acceptance.

How do I prevent anchoring bias when reviewing completed tasks?

Prevent anchoring bias during review by spawning a fresh, independent reviewer agent that validates done work against the brief without exposure to the original worker's narrative or notes, ensuring an unbiased verification outcome.

What inputs do I need to validate task completion using an independent reviewer?

Validating task completion requires the brief context, the worker's done summary, notes, and git evidence. These inputs allow the independent reviewer agent to produce a binary verdict for accepting or pruning the completed work.

Can I use a fresh agent to produce a binary verdict for work-in-progress state transitions?

Yes, you can use a fresh agent to produce a binary verdict guiding work-in-progress state transitions. The adversarial review outputs pass, needs-info, or fail states based on independent verification of the work against the brief.

When should I not use an adversarial review for task verification?

You should not use adversarial review if you lack the required inputs such as brief context, done summary, notes, or git evidence. The independent verification process depends on these elements to produce a valid binary verdict.