Group: GitHub Tikyparkinson

Sort by:Popular▼

✓Popular✓Newest

Shared model, guardrail SPI and evaluation chain for MCP agent guardrails.

Audit trail of MCP tool invocations with a pluggable audit log store.

Detects looping or compromised agents from their recent invocation history.

Holds an escalated tool invocation until a human approves or rejects it.

Declarative agent-to-tool authorization policy for MCP agents.

Detects credentials in tool arguments and redacts them from tool results.

Allowlist of outbound destinations for tools with network capability.

Rule-based prompt-injection detection over MCP tool arguments.

Fixed-window rate limiting per (agent, tool) with a pluggable counter store.

Anti tool-poisoning: TOFU fingerprint baseline of every MCP tool definition.

Escalates a session where private data, untrusted content and external comms meet.

Security and governance guardrails for MCP agents in Java.

Auto-configuration that assembles all MCP agent guardrails; the artifact end users import.