From dac491dc3fe291d4fe31af86b55c6dc79aaa828b Mon Sep 17 00:00:00 2001
From: Kerry Lu <kerry@berri.ai>
Date: Fri, 11 Sep 2026 17:30:57 -0700
Subject: [PATCH] docs(pr-template): fold Assumptions Made into Caveats

Co-Authored-By: Claude Code <noreply@anthropic.com>
---
 .github/pull_request_template.md | 36 ++++++++++++++++----------------
 1 file changed, 18 insertions(+), 18 deletions(-)

diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
index 8bdab697db4..479aac49c05 100644
--- a/.github/pull_request_template.md
+++ b/.github/pull_request_template.md
@@ -2,23 +2,6 @@
      everyday engineering language, extremely parsable and readable at a glance. This goes double for
      the TLDR, User Flow, and Caveats sections -->
 
-## Assumptions Made
-
-<!-- One bullet per claim this PR relies on that you did not actually test, each written as the claim
-     followed by what breaks if it turns out to be wrong. Write it for a reviewer who wants to attack
-     the weakest one, not to reassure them
-     Anything that narrows the scope of a bug belongs here unless you ran the test that proves it:
-     "only reproduces with X on", "no user-observable behavior difference", "this path is debug-only",
-     "no caller passes that shape". A claim you did verify is not an assumption; put the proof in
-     Screenshots / Proof of Fix instead
-     If the linked issue or ticket recorded something as untested or unverified, carry it into this
-     section or say here which run closed it out. Do not drop it silently
-     Example:
-     - The alert path is debug-only, so the growth cannot be hit at default log level. If a second
-       caller stringifies the same structure without a level gate, this ships the bug to every
-       deployment. Not tested: no run with the debug flag off
-     Write "None" if the change rests on nothing untested -->
-
 ## TLDR
 
 <!-- Fill in the bullets below and keep each one short and concrete: one line per bullet, roughly 10 words max -->
@@ -144,7 +127,24 @@ If you're seeing a delay in your PR being merged, ping the LiteLLM Team on [Slac
      - Low: anything else worth noting: naming, cleanup, an edge case nobody hits
      Nest bullets as deep as helps: hierarchy beats one long line when it makes things clearer to a
      human reader
-     Leave this section empty if there are none -->
+     Leave this section empty if there are none
+
+     Also list, under an "### Assumptions Made" subheading, one bullet per claim this PR relies on
+     that you did not actually test, each written as the claim followed by what breaks if it turns
+     out to be wrong. Write it for a reviewer who wants to attack the weakest one, not to reassure
+     them. Anything that narrows the scope of a bug belongs here unless you ran the test that proves
+     it: "only reproduces with X on", "no user-observable behavior difference", "this path is
+     debug-only", "no caller passes that shape". A claim you did verify is not an assumption; put the
+     proof in Screenshots / Proof of Fix instead. If the linked issue or ticket recorded something as
+     untested or unverified, carry it into this subsection or say here which run closed it out. Do
+     not drop it silently
+     Example:
+     ### Assumptions Made
+     - The alert path is debug-only, so the growth cannot be hit at default log level. If a second
+       caller stringifies the same structure without a level gate, this ships the bug to every
+       deployment. Not tested: no run with the debug flag off
+     Include this subheading even when the rest of Caveats is empty; write "None" under it only if
+     the change rests on nothing untested -->
 
 ## QA runbook