Prompt Injection Defense
Status: Placeholder. Content currently lives inline in enforcement. A dedicated page will extract detection and mitigation sections.
Quick Reference
Prompt injection is a user-supplied input that attempts to override the agent's system prompt or steal credentials. TnsAI provides heuristic and model-based detection.
See Enforcement for the current prompt-injection section.
Planned Content
- Heuristic patterns (role override, credential extraction, instruction smuggling).
- Model-based detection via a dedicated classifier.
- Output sanitization.
- Defense-in-depth: input filter + system prompt hardening + output review.