docs: fix incorrect claim that failed context commands produce no output

The troubleshooting page and cookbook both stated that context commands
with non-zero exit codes produce no output. In reality, ralphify captures
and injects context output regardless of exit code — this is intentional
so commands like pytest still provide useful data when tests are failing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: 2 people

docs/cookbook.md

@@ -894,7 +894,7 @@ uv run pytest --cov=src --cov-report=term-missing -q 2>/dev/null || true
 ```
 
 !!! note "Why `|| true`?"
-    Context commands that fail (non-zero exit) produce no output. Appending `|| true` ensures the coverage report is captured even when some tests are currently failing.
+    Context output is captured regardless of exit code, so `|| true` isn't strictly needed for ralphify. However, it ensures the script exits cleanly if you run it manually outside of ralphify (e.g. during debugging), and prevents errors if you later add `set -e` to the script.
 
 ### Setup commands
 

docs/troubleshooting.md

@@ -207,7 +207,9 @@ cat .ralph/contexts/git-log/CONTEXT.md
 git log --oneline -10
 ```
 
-If the command fails, its output won't be injected. Fix the command or increase its timeout.
+Context output is injected into the prompt **regardless of exit code** — even if the command exits non-zero, its stdout and stderr are still captured and included. This is intentional: commands like `pytest` often exit non-zero (because tests are failing) but still produce useful output the agent should see.
+
+If the command produces no output at all, check that it runs correctly outside of ralphify. If it times out, increase the `timeout` value in the context's frontmatter.
 
 ### Some contexts are missing from the prompt