Blog
Insights from our experts on deploying AI Agents securely and effectively)
Copilot CLI Prompt Injection Attack Chain
Copilot CLI prompt injection with encrypted payloads can leak .env files in autopilot. See the chain, two 2026 CVEs and the controls that still work.
)
)
Google ADK Security: A Production Guide
Google ADK security in production: the six 2026 CVEs to patch, callbacks, plugins, tool confirmation, MCP tool filters and A2A trust.
)
AgentCore Security: Gaps You Still Own
AgentCore security in 2026: what AWS secures, what you still own across Runtime, Gateway, Policy and Code Interpreter, and the CVEs to patch.
)
OpenClaw Skills: Setup, Picks and Risks
OpenClaw skills explained: the SKILL.md format, ClawHub install commands, a safe-install checklist and the 2026 malicious skill campaigns.
)
AGENTS.md: Guide, Risks and Security
AGENTS.md tells coding agents how to work in your repo. Learn to write one, why a planted file hijacked Codex in tests, and how to lock it down.
)
MCP Memory Server: Add Memory to Claude
Connect an MCP memory server to Claude Code with one command so your agent remembers across chats. Setup steps, file paths and security checks.
)
MCP Audit Logging: What CISOs Must Track
MCP audit logging for CISOs: what to log per tool call, how to feed your SIEM, EU AI Act, SOC 2 and ISO 42001 evidence, and shadow MCP detection.
)
Claude Code MCP and Cursor MCP Security
Claude Code MCP and Cursor MCP can run code with your credentials. Learn the config files, real CVEs and the admin controls that lock both down.
)
MCP Tool Poisoning and Rug Pull Attacks
MCP tool poisoning in 2026: how description injection, rug pulls and tool shadowing work, what new research shows, and how to stop them.
)
MCP Security Best Practices Checklist
MCP security best practices for 2026: a checklist for servers and clients on least privilege, tool pinning, sandboxing, OAuth and runtime enforcement.
)
GPT-6.1 Terra: Benchmarks & API Pricing
GPT-6.1 Terra pricing and benchmarks: $1/$5 per million tokens, independent scores vs GPT-6.1 Sol, Luna, Claude and Gemini, and the security data.
)
Claude Code Skills: The Complete Guide
Claude Code skills explained: what SKILL.md is, where skills live, how to write one, which to install and how to secure them against attacks.
)
Gemini 4 Carbon: Benchmarks & Pricing
Gemini 4 Carbon explained: benchmarks, $2/$10 pricing, who gets access, and the dual-use security risks mapped to OWASP, before access opens.
)
ChatGPT Skills: Best Picks for 2026
The best ChatGPT skills for 2026: what each does, where it wins, how to install it and how to avoid poisoned skills and supply chain risk.
)
How to Build a Company Brain
How to build a company brain in five steps: what to capture, where it lives, how to connect AI tools, who can access it and how to keep it secure.
)
OWASP MCP Top 10: Every Risk Explained
OWASP MCP Top 10 explained: the 10 Model Context Protocol risks, real attack examples and the controls that stop each one in production.
)
OWASP Top 10 for LLM Applications 2026
OWASP Top 10 for LLM Applications 2026: all 10 risks, what changed since 2025 and the enterprise controls that mitigate each one.
)
Agent Memory: A Guide to Owning Yours
Agent memory explained in plain words: what Claude and ChatGPT remember, who owns it, and how to keep your team's AI memory yours.
)
OpenClaw Alternatives: 4 Secure Options
OpenClaw alternatives compared: Hermes Agent, NanoClaw, ZeroClaw and IronClaw. See isolation, memory and security trade-offs, mapped to OWASP 2026.
)
The Institutional Memory Problem Explained
When agents do the work, where does your know-how live? The institutional memory problem explained, and how to keep it yours.
Google ADK Security: A Production Guide
Google ADK security in production: the six 2026 CVEs to patch, callbacks, plugins, tool confirmation, MCP tool filters and A2A trust.
AgentCore Security: Gaps You Still Own
AgentCore security in 2026: what AWS secures, what you still own across Runtime, Gateway, Policy and Code Interpreter, and the CVEs to patch.
OpenClaw Skills: Setup, Picks and Risks
OpenClaw skills explained: the SKILL.md format, ClawHub install commands, a safe-install checklist and the 2026 malicious skill campaigns.
AGENTS.md: Guide, Risks and Security
AGENTS.md tells coding agents how to work in your repo. Learn to write one, why a planted file hijacked Codex in tests, and how to lock it down.
MCP Memory Server: Add Memory to Claude
Connect an MCP memory server to Claude Code with one command so your agent remembers across chats. Setup steps, file paths and security checks.
MCP Audit Logging: What CISOs Must Track
MCP audit logging for CISOs: what to log per tool call, how to feed your SIEM, EU AI Act, SOC 2 and ISO 42001 evidence, and shadow MCP detection.
Claude Code MCP and Cursor MCP Security
Claude Code MCP and Cursor MCP can run code with your credentials. Learn the config files, real CVEs and the admin controls that lock both down.
MCP Tool Poisoning and Rug Pull Attacks
MCP tool poisoning in 2026: how description injection, rug pulls and tool shadowing work, what new research shows, and how to stop them.
MCP Security Best Practices Checklist
MCP security best practices for 2026: a checklist for servers and clients on least privilege, tool pinning, sandboxing, OAuth and runtime enforcement.
GPT-6.1 Terra: Benchmarks & API Pricing
GPT-6.1 Terra pricing and benchmarks: $1/$5 per million tokens, independent scores vs GPT-6.1 Sol, Luna, Claude and Gemini, and the security data.
Claude Code Skills: The Complete Guide
Claude Code skills explained: what SKILL.md is, where skills live, how to write one, which to install and how to secure them against attacks.
Gemini 4 Carbon: Benchmarks & Pricing
Gemini 4 Carbon explained: benchmarks, $2/$10 pricing, who gets access, and the dual-use security risks mapped to OWASP, before access opens.
ChatGPT Skills: Best Picks for 2026
The best ChatGPT skills for 2026: what each does, where it wins, how to install it and how to avoid poisoned skills and supply chain risk.
How to Build a Company Brain
How to build a company brain in five steps: what to capture, where it lives, how to connect AI tools, who can access it and how to keep it secure.
OWASP MCP Top 10: Every Risk Explained
OWASP MCP Top 10 explained: the 10 Model Context Protocol risks, real attack examples and the controls that stop each one in production.
OWASP Top 10 for LLM Applications 2026
OWASP Top 10 for LLM Applications 2026: all 10 risks, what changed since 2025 and the enterprise controls that mitigate each one.
Agent Memory: A Guide to Owning Yours
Agent memory explained in plain words: what Claude and ChatGPT remember, who owns it, and how to keep your team's AI memory yours.
OpenClaw Alternatives: 4 Secure Options
OpenClaw alternatives compared: Hermes Agent, NanoClaw, ZeroClaw and IronClaw. See isolation, memory and security trade-offs, mapped to OWASP 2026.
The Institutional Memory Problem Explained
When agents do the work, where does your know-how live? The institutional memory problem explained, and how to keep it yours.