Extended Threat Coverage

​

Coverage by category

​

Availability by tier

​

How detection works

• Heuristic detectors (HTML, Markdown, critic evasion, few-shot, font, memory, sub-agent, persona) use regex/pattern matching and run on every request at negligible latency cost.
• LLM-judge detectors (framing bias, RAG poisoning) use a heuristic prefilter first, then escalate to an LLM call only when the prefilter fires. This caps LLM cost to the population of suspicious requests.
• Multimodal detectors (image/audio stego, adversarial patch) operate on media attachments via the media field on the Guard API.
• Systemic correlators (sybil, fragment, cascade, collusion) run as a background service that reads from security_events, not on individual requests.

​

API integration

{
  "messages": [{"role": "user", "content": "..."}],
  "direction": "input",
  "retrieved_context": [
    {"content": "...", "source": "doc-id-123"}
  ],
  "media": [
    {"type": "image", "mime_type": "image/png", "base64": "..."}
  ]
}

​

Further reading