NeuralTrust | The leading security platform for generative AI

Application‑security detection catches code‑execution and injection patterns that ride into AI systems through prompts, tool arguments, and request bodies. It runs in‑process (no external calls) and is detection‑only — whether a match blocks is decided by the policy that references it.

Detector	Slug	Sides	Protocols	Mutable
Code Sanitation	`code_sanitation`	input	all	—

Code Sanitation — `code_sanitation`

Detects dangerous code‑injection patterns by language in the request and reports each match as a finding (signal.type: "code_injection"). It never rewrites the payload. Built‑in languages: javascript, python, php, sql, shell, html.

Settings

Field	Type	Default	Notes
`apply_all_languages`	boolean	`false`	Enable every built‑in language pattern set.
`languages`	array<`{ language, enabled }`>	—	Enable specific languages (`language` is one of the built‑ins).
`custom_patterns`	array<`{ name, pattern, description }`>	—	Your own regex signatures (`pattern` required, must compile).

{
  "name": "Flag code injection",
  "plugin_slug": "code_sanitation",
  "settings": {
    "apply_all_languages": false,
    "languages": [
      { "language": "shell", "enabled": true },
      { "language": "python", "enabled": true }
    ]
  }
}

A detector is detection‑only — you choose whether it Monitors or Blocks, and on which phase, when you add it to a policy’s Detectors tab. See Policies.

When to use

Put this in front of any system that forwards prompt/tool content into databases, shells, templating engines, or downstream APIs.
Start with the Monitor action to baseline false positives on real traffic, then promote the policy rule to Block.
It is pattern‑based and complementary to the LLM‑aware content‑security detectors — run both.

​Code Sanitation — code_sanitation

​Settings

​When to use

Code Sanitation — `code_sanitation`

Settings

When to use