โ† Back to agents
๐Ÿ“

Prompt Guard Agent

Detects and blocks at runtime prompt injections, jailbreaks and personal-data leaks, without changing your code.

What the agent does

๐Ÿšซ Anti prompt injection

Direct and indirect injections caught and blocked at runtime.

๐Ÿ”“ Anti jailbreak

Blocking of guardrail-bypass techniques: roleplay, encoding, obfuscation.

๐Ÿ”’ PII protection

Detection and masking of personal data before sending to the model.

๐Ÿ“ค Output control

Checking responses to block leaks of system information.

๐Ÿงพ Enforceable proof

SHA-256 chained audit trail of blocked attempts.

โš™๏ธ No code change

Inline deployment, no application integration.

How it works

1

Interception

Capturing the prompt before it reaches the model

2

Analysis

Running through the runtime detectors

3

Cleaning

Blocking or masking risky content

4

Proof

Attempt sealed in the audit trail

Technical approach

Runtime detectors

Semantic analysis of streamed inputs and outputs (SSE).

Indirect injection

Detection of hidden instructions in documents, RAG sources and tools.

PII masking

Entity recognition (names, emails, identifiers) before sending.

Output control

Blocking of system-prompt leaks and sensitive information.

SHA-256 chained trail

Enforceable proof of blocks.

Sovereignty

On-premise, no cloud dependency.

Use cases

CHATBOT

Public chatbot protection

Blocking injections and jailbreaks on a public-facing assistant.

INTERNAL

System-prompt leak

Preventing extraction of system instructions containing strategic information.

COMPLIANCE

GDPR masking

Masking PII in requests before processing, for GDPR compliance.

Protect your LLMs at runtime

Submit your system, we return an exposure report with the measured attack rate.