Agent♥︎Age
Catalog

GenPark Prompt Injection Canary Leak Sentinel

Official

by alphaparkinc · Python

HMAC-signed zero-width invisible canary token generator and output response leak auditor

Model Context Protocol (MCP) Server: io.github.alphaparkinc/genpark-prompt-injection-canary-leak-sentinel-skill

This MCP server generates an HMAC-signed zero-width invisible canary token and audits output responses for potential leaks. Its description indicates it is used to detect prompt-injection-related canary exposure by verifying whether protected sentinel content appears in returned output.

🛠️ Key Features

  • HMAC-signed canary token generation
  • Zero-width invisible canary tokens
  • Output response leak auditing

🚀 Use Cases

  • Detecting prompt-injection canary leaks
  • Auditing response outputs for inadvertent exposure of sentinel content

⚡ Developer Benefits

  • Provides a measurable sentinel token mechanism via HMAC signing
  • Supports developer-side verification through leak auditing of responses

Topics

agent-skillsai-safetycanary-tokenguardrailsjailbreak-defensellm-securitymcpprompt-injectionpython-standard-librarysecurity-sentinelzero-width-steganography

Related servers

More in Security