Author
Ask Claude Code to turn a release plan into sessions, items, rationale, and expected outcomes.
/test-gate:new Claude Code can write the checklist. test-gate gives you a shared place to run it.
The AI can generate a serious release checklist. Then the human has to test it by scrolling through terminal history, hunting for the current item, copy-pasting notes, and explaining findings back into chat.
Keep the human in the loop, but stop using the chat scroll as the database.
Ask Claude Code to turn a release plan into sessions, items, rationale, and expected outcomes.
/test-gate:new Open the local browser cockpit instead of reading the checklist inside the terminal transcript.
/test-gate:run Click status, add notes, paste screenshots, and promote findings without leaving the flow.
auto-save Let Claude Code digest the finished run: fail agenda, CR backlog, findings, and next step.
/test-gate:results The first internal gate replaced terminal scrolling with a single-screen verification cockpit and produced a clean, resumable artifact.
Game-Changer · Mehrwert immens hoch
test-gate is intentionally small: no cloud account, no team management surface, no CI ceremony before the schema has earned it.
The run file contains sessions, items, statuses, notes, screenshots, and findings.
Results live in the repo, so the audit trail is reviewable and reproducible.
Everything stays on your machine. No accounts, telemetry, or remote sync.
A Bun server serves the static UI and API together on localhost.
Slash commands author, run, and digest gates where the checklist came from.
The plugin gives Claude Code the author-run-results loop. The browser UI writes every status, note, finding, and screenshot back to the run JSON.
/plugin marketplace install moinsen-dev/test-gate
/plugin install test-gate@moinsen-test-gate
# then:
/test-gate:new docs/release-plans/next-release.md
/test-gate:run .test-gate/next-release.json
/test-gate:results .test-gate/next-release.json