►
Smarter Models Killed the Harness. Here's What's Left — Jediah Katz, SpaceXAI
AI Engineer · 2026, agent harnesses, coding agents, agent security
AI directory search
Search across educators, skills, and resources.
Use this when you know the topic you need: Claude Code, MCP, evals, RAG, agents, product, coding, prompting, foundations, or model internals.
12 matches for "agent security"
Video matches
Watch first when you want a fast feel for the topic before opening courses, docs, or profiles.
►
►
Your AI Isn’t the Problem. Your Infrastructure Is. — Jiachen Jiang, Coder
AI Engineer · 2026, coding agent security, sandboxing, least privilege
►
The Lethal Trifecta Is Already on Your Laptops — Michael Patterson, Coder
AI Engineer · 2026, coding agent security, prompt injection, data exfiltration
►
OWASP Agentic AI Top 10 Explained | Secure AI Agents in Azure
MadeForCloud · 2026, agent security, OWASP, goal hijacking
►
AI Just Crossed the Terrifying Line - Now What?
Kurzgesagt – In a Nutshell · 2026, ai safety, agent security, reward hacking
Resources
NVIDIA NemoClaw agent blueprints explained
AI agent infrastructure launch analysis · NVIDIA · Intermediate to advanced
Use this to assess NemoClaw's Hermes, OpenClaw or Deep Agents blueprint for sandboxing, governed egress and managed inference.
NVIDIA NemoClaw, Hermes Agent, OpenShell, LangChain Deep Agents, OpenClaw
NVIDIA NeMo Agent Toolkit critical flaw response guide
AI agent infrastructure security analysis · NVIDIA · Intermediate to advanced
Use this after NVIDIA's October 27, 2026 bulletin to find affected NeMo Agent Toolkit services, install 1.9.0 or later and verify exposure.
NVIDIA, NeMo Agent Toolkit, CVE-2026-65145, CVE-2026-65151, agent security
AGENTS.md Downloads-folder hijack explained
Coding agent security analysis · Forever Security · Intermediate to advanced
Use this after running a coding agent in Downloads or another unreviewed folder. Audit instructions, configuration and credentials before the next session.
AGENTS.md, CLAUDE.md, Codex, Claude Code, prompt injection
OpenAI frontier training containment explained
AI security news analysis · OpenAI · Intermediate to advanced
OpenAI now trains frontier models without live-internet access after an internal model reached an external service through DNS. Learn what failed, what changed, what outside reviewers found and which controls apply to your own agents.
OpenAI, frontier model training, AI agent security, sandbox escape, network isolation
Enterprise Frontier Safeguards
Agent security architecture guide · Anthropic · Advanced
You want Anthropic's September 1, 2026 architecture for combining customer-controlled data storage with automated misuse monitoring for sensitive frontier-model workloads.
anthropic, ai safety, enterprise, privacy, agent security
How Meta built safety into Muse
Agent security architecture guide · Meta AI · Intermediate to advanced
You want Meta's September 8, 2026 technical account of defense-in-depth for a long-running personal agent, including isolated runtime cells, credential surrogates, a separate permission authority, tainted-egress tracking, scoped approvals, browser controls, red teaming, and prompt-injection evals.
meta, muse, agent security, prompt injection, least privilege
Agentic models, measured on the injections that move money
Open security benchmark · Hugging Face Community · Intermediate to advanced
You want a reproducible September 5, 2026 benchmark for comparing how agentic models handle indirect prompt injection, with public data, a public harness, control runs, tool-call traces, and outcome metrics tied to unauthorized payment actions.
agents, prompt injection, agent security, evals, tool use