skills.rest
nicknisi
nicknisiCOMMUNITY
Explore all skills by this author →
Website𝕏@nicknisiGitHub

verify-behavior

Verify visible product behavior by driving real UI surfaces with checked computer-use tools.

3.0k|373|Updated Apr 12, 2012
Search Tags:#evidence#triage#computer use#QA automation#bug reproduction#acceptance testing#UI verification
One-click install
npx skills add https://github.com/nicknisi/dotfiles --skill verify-behavior
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: verify-behavior
Source: https://github.com/nicknisi/dotfiles/tree/main/home/.pi/agent/skills/verify-behavior
Command: npx skills add https://github.com/nicknisi/dotfiles --skill verify-behavior

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It replaces guesswork about whether a bug reproduces or a feature works by driving the actual UI and collecting verifiable evidence, so triage, implementation, and review decisions rest on observed behavior rather than code reading alone.

Core Features & Use Cases

  • Reproduce and verify modes: Confirm reported bugs on a baseline branch or validate implemented changes against expected behavior.
  • Checked UI interaction loop: Find roots, observe UI state, query elements, act with expect postconditions, and wait for asynchronous transitions through pi-computer-use tools.
  • Evidence-driven reporting: Aggregate verified expect verdicts, state diffs, and text extracts into triage comments, PR bodies, or review.json with explicit statuses.
  • Use Case: A reviewer needs to confirm a pull request fixes a broken checkout form; the skill launches the app, drives the form end-to-end, and reports verified postconditions with the decisive on-screen text.

Quick Start

Ask the agent to verify that the reported bug on the linked issue still reproduces on the main branch using the verify-behavior skill.

Frequently Asked Questions about verify-behavior

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

FAQPage Schema
How do I verify a UI bug fix with real interaction?▼

Use verify mode: start the app, observe the relevant UI root, drive the critical path with act_ui actions carrying expect postconditions, and quote the verified verdicts plus state diffs as evidence in the PR or review output.

How do I reproduce a reported bug before triaging it?▼

Use reproduce mode on the baseline branch, follow the path to the failure described in the issue or PRODUCT.md, and report a status of confirmed, partially confirmed, not reproduced, or blocked with observed-state evidence.

Can I use screenshots or screen recordings as verification evidence?▼

No. Evidence must come from checked tool results: expect verdicts, successor state diffs, and read_text or search_ui extracts. Building capture pipelines with ffmpeg, screencapture, or Playwright video is explicitly prohibited.

What happens if pi-computer-use is not installed or permissions are missing?▼

The skill stops with an explicit blocker instead of faking verification. Missing Accessibility or Screen Recording permissions, or an absent pi-computer-use installation, must be reported rather than worked around with osascript or xdotool.

When should behavior verification be skipped?▼

Skip it for pure backend, CI, or text-only work where no interactive surface proves the claim, or when credentials and state are unavailable. In those cases report the blocker rather than claiming verification.

Related Skills

Explore More Software Engineering Skills→
pkuppenspkuppens
1

create-validation

Generate step-by-step validation checklists with commands, URLs, and expected outcomes.

Community
Intermediate
JnusoJnuso

openclaw-pr-maintainer

Automates OpenClaw pull-request and issue triage with labeling and comment drafting.

Community
Intermediate
testdoubletestdouble
198

refactor

Refactor code iteratively with test-gated validation after each step.

Official
Advanced
badhopebadhope

e2e-test

Automate end-to-end user flow testing across Playwright, Cypress, and Selenium.

Community
Advanced