Never commit:

• OpenAI API keys
• Neo4j passwords
• .env files
• local database snapshots containing credentials

Use environment variables or the application UI for local secret injection.

Generated artifacts may contain:

• user-uploaded filenames
• runtime prompts
• model responses
• execution outputs

Treat benchmark artifacts and runtime outputs as potentially sensitive. Do not publish them casually.

If you discover a security-sensitive issue in this repository or deployment workflow, report it privately to the maintainers instead of opening a public issue with exploit details.