When a bug is found during unrelated work, file it as a ranger task
tagged `bug` instead of fixing it inline. Keeps the current task
focused and avoids losing track of the bug.

Assisted-by: Claude Opus 4.6 via pi

diff --git a/ai/skills/ranger/SKILL.md b/ai/skills/ranger/SKILL.md
index 6f566d47..be29ae9d 100644
--- a/ai/skills/ranger/SKILL.md
+++ b/ai/skills/ranger/SKILL.md
@@ -53,6 +53,7 @@ Top of the queue = most important. Bias toward quick wins — small easy tasks s
 ## Workflow
 
 - Don't mark a task **done** until the changes are committed. Commit first, then transition.
+- When you encounter a bug in ranger during other work, file it in the ranger backlog (`--backlog ranger`) and tag it `bug`. Include what you observed, the expected behavior, and how to reproduce it in the description. Don't fix it inline — continue with the original task unless the bug blocks it.
 
 ---