aegis-memory and aegis-lang
These are complementary tools where the memory/context security layer (A) would typically be implemented within or called by the security-by-construction language (B) to enforce trusted information flow in agent architectures.
About aegis-memory
quantifylabs/aegis-memory
Secure context engineering for AI agents. Content security · integrity verification · trust hierarchy · ACE patterns. Self-hosted, Apache 2.0.
Implements a 4-stage content security pipeline (injection detection, sensitive data scanning, integrity signing via HMAC-SHA256) and cryptographic agent binding to prevent context poisoning—a top attack vector in multi-agent systems. Built on an OWASP 4-tier trust hierarchy with immutable audit trails and ACE loop automation (generation→reflection→curation) that self-improves agent memory based on task outcomes. Provides Python SDK with cross-agent scoped access control and Docker-based self-hosted deployment.
About aegis-lang
RRFDunn/aegis-lang
A security-by-construction programming language for AI agents
Transpiles `.aegis` source to Python with embedded runtime security—taint tracking prevents untrusted data from reaching sensitive sinks (SQL, shell, HTML), capability decorators enforce least-privilege access, and hash-chained audit logs support regulatory compliance. Includes first-class AI agent constructs (tool invocation, multi-step plans with rollback, structured reasoning, cost budgets) and formal contract verification via Z3 SMT solver, all with zero external dependencies beyond Python 3.11+ stdlib.
Scores updated daily from GitHub, PyPI, and npm data. How scores work