Dev Tools · 20h ago
Adi-Shield v0.1.0 Detects Prompt Injection Across 5 Vectors
Adi-shield v0.1.0 is an open-source Python library that detects prompt injection attacks in five data sources: email, web, ticket, repo, and calendar. It marks untrusted data and evaluates instructions before an agent executes them, without relying on the LLM. The library passes 10 tests and is available on GitHub.
Meridian48 take
A useful defensive tool for agentic workflows, but it's still early-stage and lacks real-world deployment hooks.
Read the full reporting
adi-shield v0.1.0: detección de inyección de prompt en 5 vectores →
DEV Community
prompt-injectionai-security