Content Safety

Overview

Content Safety combines supported route-local safety plugins into one reusable policy. It is a configuration bundle, not a separate plugin type.

Key Advantages

• Reuses a consistent multi-plugin safety chain across routes.
• Keeps route-local safety readable even when several plugins are required.
• Makes the bundle explicit instead of scattering separate plugin snippets by hand.

What Problem Does It Solve?

Some routes need more than one safety control at once. The bundle keeps response screening, route-local guard prompts, and audit headers consistent across those routes.

When to Use

• a route needs several safety plugins together
• you want one reusable moderation chain for multiple routes
• the route should apply both route-local guidance and response-side screening

Configuration

Add the bundled plugin entries under routing.decisions[].plugins:

plugins:
  - type: system_prompt
    configuration:
      enabled: true
      mode: insert
      system_prompt: Apply the platform safety policy before answering and clearly note when a request needs additional review.
  - type: header_mutation
    configuration:
      add:
        - name: X-Safety-Profile
          value: standard
  - type: response_jailbreak
    configuration:
      enabled: true
      threshold: 0.8
      action: header

This is a composition example, not a content_safety plugin type. The system_prompt adds request-side guidance, header_mutation adds a policy label, and response_jailbreak evaluates the generated response. The bundle does not run a request-side content classifier, and the header is not proof that content is safe. Calibrate response screening and decide whether header-only handling is sufficient.

See the complete bundle: config/fragments/plugin/content-safety/hybrid.yaml.