01
Prompt-injection signals
Instruction patterns that could redirect an AI away from your task, alter how it handles project information or influence its output.
Standard Scan
See what .mdSiren checks inside a document.
Check supported document content and see the findings, completed checks and coverage behind each decision. Every scan record belongs to the exact version submitted.
02
Hidden and machine-readable content
Supported encoded, Unicode-manipulated, hidden or otherwise machine-readable content that deserves review.
03
Sensitive-information patterns
Supported credential and sensitive-information patterns that should be reviewed before AI use.
04
Remote links and document features
Links, active features and structural conditions that may require attention, where supported for the file format.
Format-aware
Coverage follows the file format.
PDF and DOCX can contain structural and hidden surfaces that are not present in plain-text formats. The result states what was checked for the file in front of you.
Clear evidence
Clear decisions. Explicit coverage.
A status is useful only when you can see what supports it. .mdSiren keeps the decision concise and the evidence available.
Every result is limited to the reported checks and coverage; it is not a universal safety guarantee.
- What .mdSiren found
- Which supported checks completed
- Where coverage was limited
- The recommended next step
- The destination: Approved, Needs review or Quarantine
Apply the scan
Use the same boundary across different documents.
Each guide shows how to check incoming versions, follow their decisions and retrieve current Approved copies for a practical AI workflow.
Before documents become AI context
Scan what the AI may read—not only what you can see.
Review supported findings and coverage, then retrieve the current Approved copy for your AI workflow.