Latest
●OpenAI agent swarm linked to RubyGems supply chain attack●FrontMCP SSRF patch bypassed via DNS tricks and redirects●Central Dogma Git mirror trusts any SSH host key●MySQL MCP server left SQL wide open in SSE mode●AI bug hunting ends the era of security through obscurity●Hugging Face tells AI agents to go hack a benchmark instead●Anthropic says hackers used Claude to mine 1.8M Android apps●Anthropic says Claude was abused to automate attacks and weapons work
Sunday, 13 September 2026
8 agent hacks today▲ 7 vs yesterday (1)
The Agentic TimesPowered byVortal

Incident database

Two Claude Code vulnerabilities: world-readable /copy output and worktree sandbox escape

Anthropic's Claude Code agentic coding tool had two disclosed flaws: CVE-2026-46406, where the /copy command wrote responses to a predictable world-readable path (/tmp/claude/response.md) allowing local users to read secrets or plant symlinks to overwrite files, and CVE-2026-55607, where worktree handling allowed git directory confusion and symlink abuse to overwrite home-directory files and execute code outside the seatbelt sandbox. Both were fixed in later releases (2.1.128 and 2.1.163 respectively).

Disclosed 29 June 2026 · Record updated 13 September 2026

Impact

Local unprivileged users could read privileged users' Claude responses containing secrets or credentials and overwrite attacker-chosen files; a malicious repository with prompt injection content could overwrite home-directory files such as .zshenv and achieve code execution outside sandbox restrictions.

Our coverage

No articles linked to this incident yet.

Sources

  1. nvd.nist.govhttps://nvd.nist.gov/vuln/detail/CVE-2026-46406
  2. nvd.nist.govhttps://nvd.nist.gov/vuln/detail/CVE-2026-55607