← Goodhope Technologies
Early release · free and open source

Blindspot

Finds the risky code in a pull request that no test covers. Claude reads the change and its tests together and tells you exactly which cases are missing.

The Blindspot window showing a pull request marked partly tested, with two gaps: a high-risk one about a refund larger than the amount paid and a medium-risk one about the payment provider failing part-way, each with test cases to write and a suggested test file.
Blindspot listing the untested logic in a pull request. The data shown is an example.

How it works

  1. Pick a pull requestChoose a GitHub repository and one of its pull requests.
  2. Claude reads the change and its testsBlindspot separates source files from test files, and Claude works out which new behaviour the tests exercise and which they miss.
  3. See what could break unnoticedEach gap is rated by risk and comes with concrete test cases: the situation to set up and the result to check.
  4. Take the checklist with youCopy the gaps as a checklist or post them as a comment on the pull request. Nothing is posted unless you choose to.

How it is built

  • Powered by ClaudeThe analysis is done by Anthropic's Claude models, using your own API key.
  • Honest about what it seesBlindspot reads only the lines a pull request changes and does not run your tests, so it tells you to check each gap against tests it could not see.
  • No server of oursYour keys stay on your computer. The app talks only to GitHub and to Anthropic.
  • FreeBlindspot is free to use and open source under the MIT licence. You pay Anthropic only for the runs you make.

Availability

Version 0.1.0 is an early release for Windows 10 and later, 64-bit. Expect rough edges, and please report what you find.

Download the Windows installer   Source code on GitHub

The installer is not code-signed yet, so Windows will warn that the publisher is unknown. You will need your own Anthropic API key and a GitHub access token.