aegis-shield

What problem does it solve?

Prevents prompt-injection, tool-abuse, and data-exfiltration from untrusted text by scanning it before you summarize, reply to, or write anything to memory.

Core Features & Use Cases

• Prompt-injection screening: Detects instruction-like and authority-laundering patterns that could hijack future behavior.
• Secret and exfiltration risk linting: Flags token/key/secret-like content so sensitive data is not stored.
• Safe memory append workflow: Forces a scan → lint → accept or quarantine flow, using a declarative memory entry format with source attribution.
• Use case: Before saving “web_fetch” notes from a scraped page, run Aegis Shield to quarantine high-severity content and only store sanitized, sourced memory entries.

Quick Start

Ask Aegis Shield to scan a piece of untrusted text and then decide whether to accept a sanitized memory entry or quarantine it.