Executes QA test cases in human-guided or automated-agent mode and records pass/fail results.
npx skills add https://github.com/lee-to/ai-factory --skill aif-qa-check
Please help me install this Agent Skill. Skill: aif-qa-check Source: https://github.com/lee-to/ai-factory/tree/main/skills/aif-qa-check Command: npx skills add https://github.com/lee-to/ai-factory --skill aif-qa-check
Running QA test cases manually without structure leads to lost progress, inconsistent results, and no audit trail. This Skill executes the test cases produced by /aif-qa one at a time, records pass/fail/blocked outcomes in a persistent qa-check.md artifact, and keeps results valid by tracking source, revision, and worktree digests.
Run the QA check skill in agent mode for the current branch to execute the test cases and record the results in qa-check.md.
High-intent search queries and answers about installing and using this skill.
Invoke the skill in agent mode with an optional branch argument. The agent classifies each case by execution surface, then verifies it through browser automation, CLI commands, existing test runners, API calls, or file inspection, recording evidence in qa-check.md.
Use human mode, which shows exactly one test case at a time with its steps and expected result, then asks whether it works. Your answer is recorded as Passed or Failed with your comment, and progress is saved after every case.
The skill tracks the tested commit revision, a worktree digest, and per-case source digests. When any of these change, prior results are marked Stale and unchecked, requiring retest so outdated passes are never counted as current.
Browser-dependent cases are marked Blocked if neither in-app Browser nor Playwright MCP is available, while CLI, backend-test, API, and file-inspection cases still execute. You can enable a browser capability or run the blocked cases later in human mode.
No. Destructive, irreversible, or production-side-effect actions such as database mutations, deployments, or deletions require explicit per-case user authorization. If authorization is denied, the case is marked Blocked without executing the risky action.
Agent mode cannot invent credentials, URLs, or fixtures, and cases lacking automated coverage for internal invariants stay Blocked rather than passing on weak evidence. Missing context must be supplied by the user before execution can proceed.
TravisLeeeeeeDesign QA test plans and executable test cases for features and APIs.
qa-amanAutomate QA release decisions and generate formal sign-off reports.
infantesromeroadrianTrack tool executions with usage details and error information for observability.
jsdfhasuhDefine end-to-end testing workflows with functional, performance, and security validation.
