From e50332bfff248eddeae7f41084ead3e65793823f Mon Sep 17 00:00:00 2001
From: austinried <4966622+austinried@users.noreply.github.com>
Date: Sat, 26 Sep 2026 00:30:27 +0000
Subject: [PATCH] rephrase the test bar for clarity

---
 AGENTS.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/AGENTS.md b/AGENTS.md
index a28ec197..2d30c558 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -28,7 +28,7 @@ Everything comes from the Nix flake devshell; do not install toolchains by hand.
 
 - `app/src/test` - JVM unit tests (JUnit4, Robolectric, Roborazzi, MockWebServer). Run by `:app:testDebugUnitTest`.
 - `app/src/integrationTest` - tests that need real servers. Compiled with the unit test sources but run by the dedicated `:app:integrationTest` task, and excluded from every `*UnitTest` task. The nushell harness starts the servers for you.
-- A test is only worth having if it passes for the right reason *and* fails when the behaviour breaks. There is no such thing as a documentation test: one that passes now but would not fail if the thing stopped working is actively harmful, because it hides the gap it pretends to cover. Before keeping a new test, break the behaviour (or run it against the pre-fix code) and confirm it fails, then confirm it passes again with the fix.
+- A test earns its place only if it fails when the behaviour it covers breaks, and passes when the behaviour works. A test that passes either way is worse than no test: it looks like coverage while hiding the gap. So when you add one, break the behaviour (or run the test against the pre-fix code), watch it fail, then restore the fix and watch it pass.
 
 ## CI