Claude Code secret-leak flaw fixed after Black Hat disclosure: why it is not safe to use

What happened

On August 7, 2026, *The Hacker News* reported Black Hat USA research from Novee Security about a flaw in Anthropic's coding agent tools. The report said a GitHub issue from an unprivileged account, meaning someone without special access, could reach automated CI workflows in Anthropic's coding-agent repositories.

In Claude Code, the disclosed issue was CVE-2026-54316. A CVE is a public identifier for a security bug. According to the report, the bug used Hugging Face's public download counter as an exfiltration channel, which means a way to send data out secretly. The report says this could leak an API key one character at a time. It also says every Claude Code release from 0.2.54 through 2.1.163 was affected, and that Anthropic fixed it in version 2.1.163.

What it means for you

For a normal person using an AI coding agent, the key point is simple: if untrusted content got into a Claude Code context, there was a real path for secrets to leak in default-shipped workflows. Secrets include API keys, tokens, and passwords stored for tools to use.

What to do instead

Update Claude Code to version 2.1.163 or later. Keep API keys out of projects and prompts unless they are truly needed. Treat pasted code, issue text, and repository content as untrusted unless you know where it came from. Review what workflows and tools an agent can trigger by default. Rotate any sensitive keys if you think they may have been exposed. If you use extensions or skills, prefer reviewed sources. AgentPod lists only reviewed, tested skills, which can help reduce avoidable risk.

Sources:

• https://thehackernews.com/2026/08/claude-code-and-gemini-cli-flaws-let.html?m=0