Farhan Almutairi built AL-MUNAA, a local security layer for AI agents that detects and blocks indirect prompt injections using signed HMAC fingerprints. The system combines four gates and uses GPT-5.6…
An engineer argues that prompt injection in LLMs is fundamentally a control/data boundary problem, not merely a behavior issue. They propose a two-channel architecture where untrusted content is signe…
A developer built an MCP server that connects Modal's serverless AI infrastructure to AI agents, allowing them to manage deployments, stop runaway jobs, and audit resources via natural language comman…
USTP-Secure (USTPS), a UDP-based secure transport protocol, has entered beta with authenticated packet protection, optional congestion control, and support for multiple AEAD ciphers. The protocol uses…
A developer built an OAuth shim that allows Claude Code and other agentic systems to use the Anthropic Messages API through a Claude Max subscription, eliminating per-token costs. The shim, hosted at …
A developer argues that using a boolean flag for tool-call approvals in agentic systems is insecure and proposes binding approvals to specific calls via HMAC-signed tokens. The approach prevents repla…
Easel, a shared canvas platform, gave its AI agent vision by integrating screenshots, enabling it to detect visual issues like clipped text that were invisible in the underlying JSON document. The age…