Contain direct and indirect prompt injection in LLM-integrated applications. Covers source-of-trust tagging, tool-use confirmation after untrusted input, output validation, markdown-image exfiltration prevention, and context-window hygiene. Invoke when building any app where untrusted text reaches an LLM, when the LLM has tools that act on real systems, or after a suspected injection incident.
1 email curto, só com as skills novas de Segurança. 4 minutos de leitura, sem spam, cancela com 1 clique.
Você confirma o email no primeiro envio. Sem spam. Cancela com 1 clique.
Ler descrição completaRecolher descrição
Prompt Injection Defense
Prompt injection is not a bug you can patch — it is the input/output behavior of LLMs. You can only contain it: limit what an attacker can cause to happen when they succeed, not whether they can attempt it.
This skill covers practical containment patterns for the two failure modes that actually hurt in production:
The LLM does the wrong thing — calls the wrong tool, returns the wrong data
The LLM exfiltrates — encodes secrets into an outbound URL,
[Description truncada. Veja o README completo no GitHub.]
