prismor

What problem does it solve?

Prismor addresses the security risks posed by AI coding agents by blocking dangerous actions, intercepting package installs, preventing secret leaks, and auditing session activity in real-time.

Core Features & Use Cases

• Blocking Dangerous Actions: Intercepts and blocks destructive commands, secret exfiltration attempts, and prompt injection attempts before they run.
• Intercept Package Installs: Evaluates package installations against threat intelligence before they are executed.
• Prevent Secret Leaks: Substitutes real secret values with placeholders and scrubs output to prevent sensitive information from reaching the model.
• Log and Audit: Records every tool call and maintains a tamper-evident audit trail of all agent actions.
• Discover Shadow AI: Flags AI agents running without Prismor hooks on the host machine.
• Attestation: Generates signed evidence bundles for auditors to verify the security posture and compliance of the system.

Quick Start

Install Prismor hooks into your project to enforce policy on agent tool calls:

prismor install-hooks --agent claude --workspace . --mode enforce