Bugpatrol skill
A team of AI testers that uses your app, finds bugs, fixes them in a separate worktree and opens pull requests and issues.
- 146 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Agents press buttons in the app autonomously and may change data.
- When enabled, the fixer edits code and publishing opens PRs and issues in your name.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/bugpatrolDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Agents press buttons in the app autonomously and may change data.
- When enabled, the fixer edits code and publishing opens PRs and issues in your name.
This is third-party code. Review the repository files before installing.
What it does
Bugpatrol runs a loop of four agents. An explorer uses the app like a tester and reports anything suspicious, a judge picks the real bugs and writes issues, a fixer prepares a patch in a separate git worktree, then the explorer and judge retest the fix in the running app. It supports web (Playwright), Electron, iOS and Android simulators through Maestro, HTTP APIs, CLIs and terminal UIs. Results appear in a local dashboard, in GitHub PRs and issues and in JSON files under `.bugpatrol/runs/`.
Who it is for. Small teams without dedicated QA who want a standing overnight pass over the app with ready reports.
Good fit when
- After each commit to main you want an automatic pass over the main screens.
- A mobile app is checked in a simulator without manual scripts.
- A found bug should become an issue with description and steps at once.
- A simple bugfix is handed to an agent with a retest in the running app.
Not a fit when
- The app needs strictly deterministic regression tests: the project has a separate `bugpatrol run` mode for that.
- Agents must not get repository write access or reach to the running app.
- The token budget is tight: the loop runs continuously.
Example request
Set up Bugpatrol for this repository: run the app locally, enable the patrol and leave the fixer off.Limitations
The fixer and GitHub publishing are off until switched on in `.bugpatrol/bugpatrol.yml`. A CLI agent (Claude Code, Codex, Kimi CLI, opencode, pi) or an API key is required, and every cycle spends tokens. Publishing needs `gh` with write access to the repository. If `npx` fails because of a comment in the `overrides` of `package.json`, use `pnpm dlx bugpatrol` or `npm install -g bugpatrol`. Agents press buttons in the app, so run it against test data.
How to disable. Stop the patrol process and remove the skill with `npx skills remove bugpatrol`. Close open PRs and delete the `.bugpatrol/` folder by hand.
FAQ
How do I start the patrol?
With `npx bugpatrol`. It repeats the cycle of explore, judge, fix, retest and publish until a new commit lands on main.
Where can I watch what the agents do?
`npx bugpatrol dashboard` opens a local dashboard with a live screen and status.
Can it work as a pre-merge check?
Use `bugpatrol run`: a deterministic gate for web apps with no agent, described in the documentation.
Which models work?
CLI agents Claude Code, Codex, Kimi CLI, opencode, pi, or a key for OpenRouter, Vercel AI Gateway, OpenAI, Anthropic or a custom endpoint.
What does the fixer do by default?
Nothing: the fixer and GitHub publishing are off until enabled in `.bugpatrol/bugpatrol.yml`. At first you only get reports and the dashboard.
Discussion
No questions or reviews yet.
Related
SkillsExplorbot autonomously explores a web app in the browser, finds bugs and saves flows as runnable Playwright tests.
An adversarial code-audit skill: Hunter finds bugs, Skeptic challenges them, Referee rules, and no source edit happens without your permission
A skill that tests your agent's code in a real browser: reads your changes, builds a test plan and runs it through Playwright with your own cookies
A plugin of testing skills: from requirements and test cases to API, mobile and web runs backed by traceable evidence.
The SmartBear server gives the agent BugSnag, Swagger, Zephyr, QMetry and PactFlow: errors, API specs, test cases and contracts
A self-improving QA agent: natural-language tests for web and mobile that self-heal and remember runs
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Bugpatrol (this card) | no | 146 | |
| QA Wolf CLI | no | 3,459 | runs QA Wolf scenarios; Bugpatrol explores the app itself and looks for bugs |
| Agentic Playwright | no | 283 | gives an agent Playwright control; here there is a closed loop of finding, fixing and retesting |
README in short
A five-step cycle diagram comes first, then skill installation and the agent prompt. A table lists platforms, models, GitHub and output formats, followed by links to documentation sections.