showcase: witness - evidence-first QA with real desktop browsers#1123
showcase: witness - evidence-first QA with real desktop browsers#1123texasich wants to merge 1 commit into
Conversation
| @@ -0,0 +1,80 @@ | |||
| # Witness + Executor | |||
|
|
|||
| [Witness](https://github.com/texasich/witness) is an evidence-first QA tool that drives real desktop browsers (via [cua-driver](https://github.com/trycua/cua)) through test flows, records video evidence, and reports pass/fail — all from a simple YAML spec. | |||
There was a problem hiding this comment.
The em-dash (
—) on this line violates the project's style rule in AGENTS.md, which bans em-dashes everywhere: prose, docs, code comments, and PRs. Use a comma or restructure the sentence instead.
| [Witness](https://github.com/texasich/witness) is an evidence-first QA tool that drives real desktop browsers (via [cua-driver](https://github.com/trycua/cua)) through test flows, records video evidence, and reports pass/fail — all from a simple YAML spec. | |
| [Witness](https://github.com/texasich/witness) is an evidence-first QA tool that drives real desktop browsers (via [cua-driver](https://github.com/trycua/cua)) through test flows, records video evidence, and reports pass/fail, all from a simple YAML spec. |
Context Used: AGENTS.md (source)
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| target: https://myapp.com | ||
| browser: chrome | ||
| steps: | ||
| # Use Executor to reset test state via API | ||
| - action: executor_call | ||
| tool: "myapp resetTestData" | ||
| args: '{"user": "test@example.com"}' | ||
|
|
||
| # Then drive the real browser | ||
| - action: navigate |
There was a problem hiding this comment.
executor_call presented as an implemented action
The YAML block shows action: executor_call as a working step, but the PR description describes this integration as aspirational ("Witness could use Executor as its tool backend"). A reader may try this spec and get a runtime error when the action is unrecognised. The Status section at the bottom lists no ✅ for Executor integration, which is easy to miss. Consider adding a note like # planned above the executor_call step, or moving it under a separate "Planned integrations" heading, so the capability boundary is unambiguous.
Witness — an open-source QA tool that drives real Chrome through test flows, records video evidence, and reports pass/fail.
Built in response to your tweet asking for an "OpenDevin for autonomous QA."
How it works
Test spec (TodoMVC — 8/8 passed)
Key differences
Witness + Executor
Witness could use Executor as its tool backend for API/GraphQL/MCP calls during test flows.
Status
Would love your feedback — happy to jump on a call.