dogfood

The dogfood skill is designed to support systematic exploratory quality assurance testing for mobile applications on both iOS and Android using the agent-device tooling environment. It helps identify runtime issues such as bugs, usability problems, broken workflows, and unexpected behaviors by interacting directly with the app rather than analyzing source code. The skill provides a structured workflow for opening test sessions, capturing screenshots and snapshots, exploring application flows, collecting issue evidence, and closing sessions in an organized manner. It is intended for real-device or agent-driven runtime validation where findings are based entirely on observed app behavior.

The skill includes safeguards and environment validation steps to ensure the testing setup is reliable before execution begins. It requires access to a trusted agent-device binary and verifies that the installed CLI version meets the minimum supported requirement of 0.14.0. The workflow emphasizes using official CLI guidance through the built-in help system to obtain the latest reporting formats, evidence collection commands, and testing recommendations. Tool access is intentionally restricted to agent-device and npx agent-device commands, helping maintain a controlled and focused QA environment. The testing loop includes session management, snapshot capture, visual documentation, flow exploration, evidence gathering, and issue reporting.

This skill is useful for QA engineers, mobile developers, product teams, and AI-assisted testing workflows that require exploratory testing of mobile applications. It is particularly valuable during pre-release validation, regression testing, usability reviews, and bug-hunting sessions. Teams can use it to validate real user experiences across supported mobile platforms while ensuring that reported findings are grounded in observable runtime behavior rather than assumptions or static code inspection.