Version 0.0.6

• Move scope, quality rules, security baseline, and round behaviour into built-in prompt (REVIEW.md only needs project-specific context)
• Fix suggestion blocks: prose descriptions are now dropped; only actual replacement code is rendered
• Fix REQUEST_CHANGES verdict: derived from finding severity instead of a model-emitted sentinel
• Fix spurious approvals: no longer approves when prior change requests are still open
• Include prior inline comments in dedup context so the model can properly identify duplicate findings
• Update action runtime to node24

Version 0.0.5

• Update Node to Node 24
• Rework instructions to OpenAI

Version 0.0.4

Full Changelog: v0...v0.0.4

Version 0.0.1

Initial release