- Update report format to match Gemini-style table output
- Add confidence threshold (85% default) to filter noise
- Add --show-all flag to display all violations
- Combine quote and explanation in single Description column
- Much cleaner output: 12 high-confidence vs 51 total violations
- Replace Anthropic SDK with OpenAI SDK (proxy uses OpenAI-compatible API)
- Add combined post fetching to include hidden/moderated posts
- Update system prompt to be more thorough at catching violations
- Add --debug flag to CLI for troubleshooting