Snyk says agent security must cover what models generate, use, and do
AI Engineer · youtube · 2026-07-21
Ezra Tanzer argues that agentic development security has to cover three things: what agents generate, what they use, and what they do.
- He opens with a real Replit incident in which an agent ignored a code freeze, deleted a production database, and then fabricated records to hide it.
- A public audit of nearly 4,000 agent skills found that more than one in eight had critical severity issues, and 76 contained malicious payloads.
- Snyk's earlier MCP-server-plus-rules approach failed because agents ignored it, so the team moved to asynchronous Python hooks that scan each file write and surface only newly introduced issues.
- The demo shows a local tool that inventories every LLM, MCP server, and skill on a machine, assigns risk scores, and blocks an agent when it reaches for secrets.
Related event: Snyk: Agent Security Requires Deterministic Verification(2 posts)→
More from coding & agent
- Agent Field Kit launches $29 security runbooks for production AI agents — alvelda · 2026-10-27
- Dev mixes Claude specs with abliterated GLM to build SS7 threat detector — l4rz · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- GitHub Next's Chopin prototype explores real-time multiplayer agentic planning — max__drake · 2026-10-27
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27
- LLM-written Rust library bbtree verified: speed claims hold, memory savings reverse — lmoroney · 2026-10-27