feat: improve AI reviewer assumption handling to reduce spam

Update the prompt to prevent the AI from spamming users with questions
about unverifiable issues. Changes:
- Default to skipping issues that can't be verified from the diff
- Only ask for clarification on critical issues (security, breaking bugs, data loss)
- Require explicit labeling of inferences as "Inference (not verified)"
- Emphasize focusing on directly visible issues in the diff

This makes reviews more honest and actionable while avoiding noise
from assumptions about code not shown in the diff.

Author: LearningCircuit

ai-reviewer.sh

@@ -218,7 +218,11 @@ PROMPT="You are an expert code reviewer. Please analyze this code diff and provi
 
 Focus on security, performance, code quality, and best practices.
 
-Focus on high-value issues. Style suggestions are welcome if impactful, but not minor optimizations. Be concise and dense - use bullet points for clear structure. Avoid repetition - in summary sections, only repeat critical issues (security, bugs, breaking changes). If flagging issues about code not visible in the diff, clearly state what you're assuming and why. For non-critical improvements, consider approving with recommendations rather than requesting changes.
+Focus on high-value issues. Style suggestions are welcome if impactful, but not minor optimizations. Be concise and dense - use bullet points for clear structure. Avoid repetition - in summary sections, only repeat critical issues (security, bugs, breaking changes). Important: Focus on issues directly visible in the diff. If you cannot verify something from the diff alone (e.g., missing context, unclear defaults, code not shown):
+- Default: Skip the issue to avoid spam
+- Only ask for clarification if it's critical (security vulnerabilities, breaking bugs, data loss risks): \"Cannot verify [X] from diff - please confirm [specific question]\"
+- If making an inference about non-critical issues, explicitly label it: \"Inference (not verified): [observation]\"
+For non-critical improvements, consider approving with recommendations rather than requesting changes.
 
 Required JSON format:
 {