NIBVOK AI Security
A runtime security enforcement layer for AI agents that intercepts commands and applies a three-state allow/confirm/deny decision framework to prevent agents from executing destructive operations. It supports governance scenarios including secret key protection, database security, and git operation controls.
Timeline (last 90 days)
NIBVOK AI Security v0.1.2 is a runtime enforcement layer for OpenClaw agents.
NIBVOK's core claim is to read the specific commands a tool is about to execute, rather than matching tool names.
NIBVOK AI Security returns one of three answers—allow, confirm, or deny—before action execution for every tool call.
High-risk behaviours governed by NIBVOK v0.1.2 include destructive deletion, credential/key access leakage, writes outside the workspace, destructive git operations, database deletion, transactions equal to or exceeding $500, modifications to the governance layer itself, and outbound message sending.
NIBVOK's confirm state explicitly transfers decision-making authority to a human operator within the action window, drawing on the Human-in-the-Loop (HITL) design pattern.
NIBVOK's deny state is designed to be irreversible, allowing no retry or downgrade path to bypass a denial decision.
NIBVOK ships with a public test suite and a public incident log (INCIDENTS.md) that records four real bypass vulnerabilities discovered and fixed during development: an overly broad write rule, misclassification of read operations as write operations, string-anchor-based key matching, and path traversal.
The NIBVOK release artifact is accompanied by the SHA-256 checksum 3c05f7d226f46a6c201a4613e1f9275ccc2b858f76588068ad1a154f970dc3bc.
The package @nibvok-llc/nibvok-ai-security has been listed on the ClawHub registry.
NIBVOK's product team states that this layer cannot guarantee 100% data protection and is only an additional layer of defence.
All Facts (10)
NIBVOK's product team states that this layer cannot guarantee 100% data protection and is only an additional layer of defence.
50%UnverifiedNIBVOK's core claim is to read the specific commands a tool is about to execute, rather than matching tool names.
50%UnverifiedNIBVOK AI Security returns one of three answers—allow, confirm, or deny—before action execution for every tool call.
50%UnverifiedHigh-risk behaviours governed by NIBVOK v0.1.2 include destructive deletion, credential/key access leakage, writes outside the workspace, destructive git operations, database deletion, transactions equal to or exceeding $500, modifications to the governance layer itself, and outbound message sending.
50%UnverifiedNIBVOK's confirm state explicitly transfers decision-making authority to a human operator within the action window, drawing on the Human-in-the-Loop (HITL) design pattern.
50%UnverifiedNIBVOK's deny state is designed to be irreversible, allowing no retry or downgrade path to bypass a denial decision.
50%UnverifiedNIBVOK ships with a public test suite and a public incident log (INCIDENTS.md) that records four real bypass vulnerabilities discovered and fixed during development: an overly broad write rule, misclassification of read operations as write operations, string-anchor-based key matching, and path traversal.
50%UnverifiedThe NIBVOK release artifact is accompanied by the SHA-256 checksum 3c05f7d226f46a6c201a4613e1f9275ccc2b858f76588068ad1a154f970dc3bc.
50%UnverifiedNIBVOK AI Security v0.1.2 is a runtime enforcement layer for OpenClaw agents.
50%UnverifiedThe package @nibvok-llc/nibvok-ai-security has been listed on the ClawHub registry.
50%