io.github.Hiroshi-Ichiyanagi/kagami MCP Server
The MCP server io.github.Hiroshi-Ichiyanagi/kagami provides a deterministic prompt-injection detector. It is described as signed and capable of producing offline-verifiable verdicts. The project is explicitly noted as not an LLM, focusing on detection rather than text generation.
🛠️ Key Features
- Deterministic prompt-injection detection
- Signed verdicts
- Offline-verifiable outcomes
- Not an LLM
🚀 Use Cases
- Checking whether prompts include signs of prompt injection
- Enabling verification of detection results without network dependencies
⚡ Developer Benefits
- Deterministic (repeatable) detection behavior
- Verification of verdicts via offline methods
- Clear non-LLM scope for security-focused workflows
⚠️ Limitations
- Limited to prompt-injection detection; not designed for general language model tasks