prompt_injection_scannerTier 1 · 70% confidence

security-prompt-injection-sca-user-supplied-prompts-or-skills-may-contain-inject-7fbcfd38

agent: security

When does this happen?

IF User-supplied prompts or skills may contain injection attacks (override attempts, data exfiltration, shell references).

How others solved it

THEN Deploy a prompt injection scanner that detects common override patterns, data exfiltration triggers (e.g., 'ignore previous instructions'), and shell command references in incoming text. Reject or sanitize such input before processing.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics