If you discover a security vulnerability in this project, please report it responsibly.

Do NOT open a public GitHub issue for security vulnerabilities.

Please report security issues via the SAP Trust Center at https://www.sap.com/about/trust-center/security/incident-management.html.

Alternatively, you can email secure@sap.com.

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

• Acknowledgment: Within 48 hours
• Initial assessment: Within 5 business days
• Resolution target: Varies by severity

This toolkit follows defense-in-depth principles:

1. Credentials never reach the LLM — The MCP server holds secrets in process memory; only sanitized responses flow to the agent.
2. Application credentials over passwords — Scoped, revocable, auditable.
3. Keychain storage — Secrets retrieved via OS commands (security find-generic-password on macOS, pass on Linux), never stored in plaintext config.
4. Response sanitization — The MCP server strips sensitive fields before returning results.
5. Project-scoped access — All operations are scoped to the authenticated project.
6. Audit trail — All actions are logged to Hermes with the credential identity.

This policy covers:

• The agent toolkit (skills, rules, knowledge)
• Plugin manifests and configuration
• The validation tooling

For vulnerabilities in the MCP server itself, please report to the openstack-mcp-server repository.