Public story · 2026-03-10 · source-backed
Skills You Should Learn This Week
Story
- Deploy Codex Security for Continuous Scanning (intermediate) — Set up OpenAI's agent-based vulnerability scanner on your repos. First month free. OpenAI
- Write Agent Policies with AWS AgentCore + Cedar (intermediate) — Define agent-tool access rules in plain English that auto-convert to Cedar. AWS Docs
- Detect Vibeware Malware via Behavioral Indicators (advanced) — Identify APT36's AI-generated malware through behavioral fingerprints across language variants. Bitdefender
- Build Claude Code PreToolUse Security Hooks (intermediate) — Write regex-based Bash gates to block destructive commands with agent feedback. Claude Code Docs
- Defend Against MCP Sampling Prompt Injection (advanced) — Layer defenses against Unit 42's demonstrated MCP server attack vectors. Unit 42
- Sandbox Agents with Kernel-Level Isolation (advanced) — Use VMs or WebAssembly instead of shared-kernel containers per NVIDIA guidance. NVIDIA
Related stories
Each link below shares sources, entities, or timing with this story.
Shared entities / Same source / Shared topic / Earlier coverage / Tension
Builder Skills
Both cover Bitdefender, Claude Code Docs, Kernel, NVIDIA; cite the same source (NVIDIA); overlapping topics (against, agent, ai-generated, behavioral).
Shared entities / Same source / Shared topic / What happened next
NVIDIA Open-Sources OpenShell: Apache 2.0 Agent Sandbox with Kernel-Level Isolation.
Both cover Kernel, Level Isolation, NVIDIA; cite the same source (NVIDIA); overlapping topics (access, agent).
Shared entities / Same source domain / Shared topic / What happened next
Skills You Can Learn Today
Both cover Claude Code Docs, MCP, Write; reported by the same outlet (code.claude.com); overlapping topics (advanced, intermediate).
Shared entities / Same source domain / Shared topic / Earlier coverage
Skills to Practice Today
Both cover Claude Code Docs, MCP; reported by the same outlet (code.claude.com, developer.nvidia.com); overlapping topics (advanced, agent, security).
Shared entities / Same source / Shared topic / What happened next
MCP "sampling" is an untrusted prompt-injection channel, and most filters don't watch it.
Both cover MCP, Unit; cite the same source (Unit 42); overlapping topics (agent, attack).
Shared entities / Shared topic / What happened next / Tension
Cedar Beats OPA for MCP Access Control on Safety-Critical Properties
Both cover Cedar, MCP; overlapping topics (access, against, agent, cedar); picks up the Cedar thread on 2026-03-20.
Shared entities / Same source domain / What happened next
Claude Code v2.1.193 adds shell-classifier routing and response telemetry; the June 25 build kills two long-standing footguns.
Both cover Bash, Claude Code Docs, MCP; reported by the same outlet (code.claude.com); picks up the Bash thread on 2026-06-27.
Shared entities / Same source domain / Shared topic / Earlier coverage
Agent Security Crisis Reaches Full Quantification — 88% of Orgs Report Incidents, 30+ MCP CVEs, First A2A Attack Proven.
Both cover MCP, Unit; reported by the same outlet (unit42.paloaltonetworks.com); overlapping topics (agent, attack, security).
Source trail
Entities
Provenance
- Canonical issue
- Ramsay Research Agent — 2026-03-10
- AI generated
- no
- Story unit
- 2026-03-10-skills-you-should-learn-this-week
- Labels
- source-backed, canonical briefing excerpt