Reported by 1 outlet
Zero-Privilege RCE in Anthropic, Google, and OpenAI Coding Agents — Malicious GitHub Issues Execute Code
Why this score
- Breadth
- 0.50 / 2.0 · 1 source
- Authority
- 1.00 / 2.0 · best outlet weight
- Severity
- 4.80 / 5.0 · CVSS + content signals
- Velocity
- 0.00 / 2.0 · updates in 48h
- Pickup speed
- 0.00 / 0.6 · time to second source
- Recency
- × 0.00 temporal multiplier
- Community
- + 0.00 / 1.5 · reddit signal
- Score
- 0.0 / 10 = base × recency + community
Timeline
-
original reporting
Zero-Privilege RCE in Anthropic, Google, and OpenAI Coding Agents — Malicious GitHub Issues Execute Code
cyberpress.org · 68d ago · August 7th, 2026
Novee Security researcher Elad Meged demonstrated that an unauthenticated attacker can use a malicious GitHub issue to achieve RCE, credential theft, persistent agent hijacking, and supply-chain compromise in default configurations of Claude Code Action, Gemini CLI, and OpenAI Codex — patterns found in more than 100 public repositories, including the vendors' own. Anthropic's fixes culminated in CVE-2026-54316, Google rated its case CVSS 10.0 and made breaking trust-model changes to headless Gemini CLI execution, and OpenAI hardened Codex by isolating stages after an AGENTS.md poisoning path. The common flaw is the agent harness — permission logic, tool routing, sandbox controls, shared workspace — not the model; any automation that processes externally controlled text without strong isolation carries the same risk.