prompt injection
3 posts tagged prompt injection. Atom feed for this tag.
- Why does Agentforce take orders from a web form? To the model, a stranger's lead and the employee's question are one stream of text. The filter meant to stop the leak only blocked URLs it recognized.
- If the model refused the attack, why isn't the server safe? A refusal is the model's call, and the model's call can change. mcploitable is a lab of deliberately vulnerable MCP servers built to show where that stops mattering: 105 attempts at the correct control, zero through.
- Why does my MCP server call the tool I told it never to call? The system prompt is advice to the model. Nothing between the model and the API ever reads it, so the checks have to live in code.