Choosing the wrong testing interface can increase effort and leave acceptance requirements unverified. This skill selects suitable interfaces and checks observable outcomes with scoped desktop helpers.
Review the Skillstore skill "software-testing-router" from https://skillstore.io/skills/hunxuankai-software-testing-router.md and its manifest at https://skillstore.io/api/skills/hunxuankai-software-testing-router/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.
エージェントは引き続き計画を提示し、セキュリティポリシーで必要な確認を求める必要があります。
エージェントが読めるリソース
AI エージェント、クローラー、スクリプトがページ全体ではなく整理されたコンテキストを必要とする場合は、これらのリンクを使ってください。
All 152 static findings are false positives involving documented commands, scoped automation, local evidence files, research URLs, and multilingual text. No evidence found of credential theft, data exfiltration, malicious obfuscation, or prompt injection. Desktop actions can modify the selected application; macOS native behavior remains unverified on actual hardware.
hunxuankai. (2026). software-testing-router security audit report (audit version 1) [Author version 1.1.1]. Skillstore. https://skillstore.io/skills/hunxuankai-software-testing-router/audits/1
BibTeX形式の引用
@techreport{hunxuankai-hunxuankai-software-testing-router-2026,
author = {hunxuankai},
title = {software-testing-router security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/hunxuankai-software-testing-router/audits/1},
note = {Author version 1.1.1}
}
CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "software-testing-router security audit report (audit version 1)"
version: "1.1.1"
type: report
authors:
- name: "hunxuankai"
date-released: "2026-11-08"
url: "https://skillstore.io/skills/hunxuankai-software-testing-router/audits/1"
identifiers:
- type: other
value: "skillstore:hunxuankai-software-testing-router:audit:1"
description: "Skillstore immutable audit report identifier"
Skillstore スコア
このスコアの理由証拠の信頼度: 中
73
アーキテクチャ
100
保守性
87
コンテンツ
65
コミュニティ
100
仕様準拠
作成できるもの
Validate a Packaged Windows Export
Inspect an EXE, fill its form, invoke export, and independently verify fresh output without requiring application source.
Check Browser Submission Behavior
Verify that a real button click produces the expected request, business response, interface state, and output artifact.
Separate Qt and Delivery Checks
Plan signal and cancellation tests separately from packaged application and visual acceptance checks.
これらのプロンプトを試す
Choose a Starting Interface
Review [application] and [acceptance requirement]. Identify available testing interfaces, recommend a minimal first check, and state what remains unverified.
Inspect a Windows Application
Inspect [Windows EXE] using its observed PID and executable path. Report windows, supported UIA patterns, unique selectors, and blockers without changing application state.
Verify a Browser Export
Test [URL] using real form interaction and [expected output]. Correlate the click with its response, verify business results, and inspect fresh output.
Handle Interrupted Desktop Actions
Review [desktop workflow], [journal], and current output artifacts. Distinguish completed steps from unknown actions, avoid automatic replay, and propose targeted verification.
ベストプラクティス
Select the smallest testing interface that can prove the acceptance requirement, and document its verification boundary.
Confirm target identity and unique control scope before actions; use condition waits and a bounded worker deadline.
Use fresh output directories or task identifiers, verify actual artifact contents, and inspect journals before recovering interrupted actions.
回避
Treating API success, a returned button invocation, or an existing file as proof of a complete user workflow.
Hiding ambiguous selectors with the first match, guessing coordinates, or automatically replaying actions after a timeout.
Claiming unsupported platform coverage, counting skipped tests as passed, or promising an unmeasured speed improvement.
よくある質問
Does this skill require application source code?
Windows UIA can inspect suitable packaged applications without source. Qt internal tests require source and a compatible project environment.
Which systems have execution evidence?
The package records Windows 10 examples. Native macOS, Windows 11, and Linux execution are not established by those results.
Does it install testing dependencies automatically?
The runtime package does not include development dependencies. Use existing project tools and obtain authorization for any required installation.
Can it change Accessibility permissions or elevate privileges?
No. The macOS helper checks existing permission without prompting or modifying TCC, and the workflow does not authorize elevation.
What happens when a desktop action times out?
The worker stops, while the target application remains running. Review the journal and actual state because a started action may have completed.
Does it replace screenshots and visual acceptance?
No. Layout, clipping, colors, and rendering require controlled visual evidence; successful API or accessibility checks cannot prove appearance.