One guardrail layer between any agent and the LLM.
Trishul is a pluggable middleware proxy that enforces declarative safety policy across the whole agent lifecycle — inbound prompt-injection & PII firewall, policy-aware tool execution with human-in-the-loop approval, and a streaming egress verifier that catches secrets split across token boundaries. No changes to your app code.
Ingress firewallTool-policy HITL
Streaming egress verifierPolicy as code
Circuit breaker
Ingress firewall
Detect prompt injection & mask PII / secrets before the prompt reaches the model.
Try:
injection
PII bundle
leaked key
How it decides
Two-layer detection, one declarative policy.
1 · Deterministic — regex+checksum PII (Luhn card, Verhoeff Aadhaar), 15+ secret types, weighted injection signals.
2 · LLM-judge — only ambiguous injection scores escalate to the model, so you pay per-call rarely.
3 · Policy — thresholds, mask modes & block severities all live in policy.json.
4 · Large context — sliding window scores injection buried deep in long inputs.
2 · LLM-judge — only ambiguous injection scores escalate to the model, so you pay per-call rarely.
3 · Policy — thresholds, mask modes & block severities all live in policy.json.
4 · Large context — sliding window scores injection buried deep in long inputs.
Proxy output appears here.