feat: Force exhaustive first-pass analysis in code review prompts

[alex-alecu]

Root cause

AI agent was interleaving analysis and commenting — it would find an issue, post it, then move on. The new prompt enforces a strict two-phase approach: complete all analysis first, then post everything at once. This should eliminate the drip-feed pattern across revisions.

The fix

1. New hard constraint App Builder - Unit and Integration tests #7 — "ALL issues in ONE pass": Explicitly tells the AI to report every issue it finds in a single review and frames finding new issues on unchanged code in subsequent revisions as a failure of the first review.

2. Restructured workflow — Two key changes:

   • Step 1 renamed from "Analyze the PR/MR" to "Analyze ALL Changed Files (complete this BEFORE posting any comments)" with a bold instruction: "Do NOT post any comments until you have reviewed EVERY changed file. Analyze ALL files first, THEN comment."
   • The per-file instructions now explicitly say "Check for ALL issue types" and "Note every issue you find — do NOT stop at the first issue per file"
   • Steps 2-3 are renamed to emphasize "ALL Issues" rather than individual verification

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (2 files)

• src/lib/code-reviews/prompts/default-prompt-template.json
• src/lib/code-reviews/prompts/default-prompt-template-gitlab.json