Building 70 Projects ranging from beginner to advanced so anyone can β learn from, build upon, use as a reference, or even copy directly. Owned by my cybersecurity company π
Go aicertificationcloud-engineeringcybercybersecuritycybersecurity-analystcybersecurity-certificationcybersecurity-certificationscybersecurity-educationcybersecurity-jobscybersecurity-portfoliocybersecurity-projectscybersecurity-toolsgrchackinglearningpentestingproject-repositoryprojetcspython
Update README.md
Carter Perez4cabd058bad2
diff --git a/PROJECTS/beginner/prompt-injection-firewall/README.md b/PROJECTS/beginner/prompt-injection-firewall/README.mdindex db57f802..e5eab413 100644--- a/PROJECTS/beginner/prompt-injection-firewall/README.md+++ b/PROJECTS/beginner/prompt-injection-firewall/README.md@@ -20,6 +20,9 @@[](#build-and-test)[](https://www.gnu.org/licenses/agpl-3.0)+---+[](https://notsandboxed.com)+> A prompt injection firewall that gives up on reading the attacker's mind and enforces structure around the model instead. Untrusted text is fenced behind a per-request nonce it cannot guess, the model's tool calls are authorized by the firewall rather than requested-and-granted, and registered secrets are matched on the way out through a closure over every reversible encoding. Text inspection ships too, scored and labelled best-effort, because it catches the copy-pasted eighty percent. It runs keyless and offline, ships an OpenAI-compatible proxy, and includes a six-level arena where each level is a firewall configuration so you can feel which layer stopped you.## Why ingress filtering cannot be the answer