
1/ JetBrains just published 7 new CVEs for its Junie AI coding agent (GitHub Action, CLI and IDE plugin). If you use Junie, here's what to update 🧵 2/ The two rated high: • CVE-2026-133198 (7.8): code execution via terminal environment settings in project files (plugin before 252.819.35 / 253.819.35) • CVE-2026-133195 (7.5): Junie CLI's file tools could modify files outside the project (before 26.6.29) 3/ The Junie GitHub Action before 1.10.0 had three more: untrusted fork code processed in privileged runs, path traversal in feedback session markers, and command injection via the version input. Plus two more CLI bugs (overwrite via shell redirection, command injection via file names in code review). 4/ Fix: Action → 1.10.0+, Junie CLI → 26.11.1+, IDE plugin → 252.819.35 / 253.819.35+. And be careful opening untrusted repos with an agent: two of these trigger from project files or file names. https://vibevulns.com/issues/jetbrains-junie-seven-cves-github-action-fork-code-command-injection-cli-file-writes-terminal-env-rce-cve-2026-133192-133198/ 5/ Same week: 5 more malicious npm packages posing as AI SDKs and MCP servers (character-ai-sdk, char-ai-sdk, @mibbp/agent-observer, mrnima-openai-chatgpt-bot, @fdcasj_dfew/md-doc-mcp). If one ran via npx, treat the machine as compromised and rotate keys. https://vibevulns.com/issues/malicious-npm-ai-sdk-and-mcp-packages-character-ai-sdk-char-ai-sdk-agent-observer-md-doc-mcp-2026-11-19/
