Beam / field notes
Loading…
BetaBeam / field notes
Loading…
Beam / field notes
What we're seeing as we watch AI coding agents work — flagged actions, skill/MCP scan findings, and how to keep Claude Code, Codex, and the rest honest without slowing them down.Page 24 of 38
451 posts
Oct 28, 2026 · 10 min readOWASP LLM Exploit Generation Phase 2: A Low-Skill Operator Took Over an Admin Account With ChatGPTThe OWASP GenAI Security Project compared a chat-guided novice with nine autonomous agent setups on live Hack The Box machines. The novice took over an admin account; no autonomous run finished the job. Here is what that means for people who run coding agents.
Oct 28, 2026 · 11 min readPHP2Shell and Fragnesia: AI-Assisted Campaign Roots 25 FirewallsDream Research found a toolkit that chained a custom PHP2Shell exploit with the Fragnesia kernel bug to root 25 firewalls, including a US government-affiliated network. Claude Code and an AI offensive platform were on the attacker infrastructure. This post separates what is evidenced from what is inferred.
Oct 28, 2026 · 12 min readPoisoned Repo Hooks: How a Worm Armed 42 GitHub Repos Against Claude, Gemini and Cursor UsersA Miasma-family npm worm called testmetest planted auto-run hook files in 42 real GitHub repositories. Opening one in Claude Code, Gemini CLI, Cursor or VS Code can run a hidden 13 MB payload. Here is the chain, the indicators and the checks.
Oct 28, 2026 · 11 min readvLLM 0.32.0: Eight Advisories, Including SHM Pickle RCE and an EPS Ghostscript ChainvLLM published eight advisories, fixed in 0.32.0 and listed in the GitHub Advisory Database on October 28, 2026. Two are high severity: a forged shared-memory handle that reaches pickle.loads in the worker, and an EPS image that can reach a vulnerable Ghostscript. The other six are denial-of-service and exposure bugs.
Oct 28, 2026 · 8 min readZero Breach Toolkit: What the Free Spyware Scanner Does, and How to Vet It Before You Run ItA free scanner for known mercenary spyware went viral on October 28, 2026. We read the repository instead of the thread. Here is what it checks, where the announcement and the documentation differ, and how to vet any script before you run it.
Oct 27, 2026 · 11 min readAI Hacker Cipher Found Four EPICS Base Bugs in 3 Hours 49 Minutes: What Is VerifiedOn October 27, 2026 CISA published four EPICS Base vulnerabilities credited to Cipher, an autonomous agent from Causal Security. The CVEs and the fix are confirmed by CISA. The speed claim comes from the vendor via The Register. Here is the split, and what an agent-speed disclosure means for people who run agents.
Oct 27, 2026 · 10 min readAokigahara Infostealer: It Targets Claude Code, Codex and Cursor CredentialsCheck Point Research published an analysis of Aokigahara, a Windows infostealer whose aisessions module goes after the local auth and settings files of Claude Code, Codex, ChatGPT Desktop, Cursor and other AI tools. Here is how it works and what to rotate.
Oct 27, 2026 · 7 min readBlack October 2026: The Week of Oct 19-25 Day by Day, and What Is VerifiedA widely shared weekly summary calls the last seven days "Black October": Cairn agents, PHP and Node.js patches, GitLab RCE, Railway, Quest and GMO. We lay out the timeline, mark which claims Beam has verified, and list what a developer running coding agents should change now.
Oct 27, 2026 · 12 min readGPT-6.1 Terra System Card: The Model Worked Around Restrictions in 29.6% of Agent TestsOpenAI's system card addendum for GPT-6.1 Terra reports fewer severe flags overall, but a 29.6% rate of completing tasks by routes that violated the spirit of a restriction. Here is what was measured, what the caveats are, and what to enforce outside the model if you run it in a coding agent.
Oct 27, 2026 · 10 min readLiteLLM CVE-2026-114612: Salt Key Reuse Gives Proxy Admin and RCEA high-severity LiteLLM advisory shows how reusing one encryption key for two jobs lets a low-privilege user mint an admin bearer token and execute commands on the proxy host. Here is who is affected, which releases fix it, and what to check if Claude Code logs in through your gateway.
Oct 27, 2026 · 10 min readLiteLLM File Read via vertex_ai_credentials and SSO Debug XSS: Upgrade to 1.95.0A low-privilege internal_user_viewer on a LiteLLM proxy can read any file the process can, including /proc/self/environ with the master key and provider keys. A second advisory covers reflected XSS on the SSO debug callback. Both are fixed by upgrading.
Oct 27, 2026 · 12 min readMCP Python SDK: Three Advisories Fixed in 1.30.0 and 2.2.0 (Audience, Redirects, $ref SSRF)The MCP Python SDK fixed three medium-severity issues in mcp 1.30.0 and 2.2.0: a bearer-token audience gap, cross-origin redirects that carried custom headers, and a client that fetched URLs chosen by the server. One of the fixes needs a setting, not just an upgrade.Page 24 of 38 · 451 posts