Socket reports a new wave of the self-spreading Shai-Hulud supply-chain worm, in its Miasma and Hades variants, that compromised more npm packages and, for the first time, reached the Go ecosystem. On June 24 attackers used a hijacked maintainer account to push trojanized versions of LeoPlatform and RStreams npm packages, tied to cloud and serverless workloads, and also poisoned a Go module from the Verana blockchain project. The malware harvests developer and CI/CD credentials, abuses GitHub Actions, and polls GitHub hourly for a marker commit to pull down its Hades payload. Researchers note the campaign keeps shifting ecosystems and indicators to stay ahead of detection rather than changing its core behavior.
Wiz Research found a high-severity flaw in Amazon Q Developer, Amazon's AI coding assistant, that let a malicious code repository run commands and steal a developer's cloud credentials simply by being opened. The bug (CVE-2026-12957) lay in how Amazon Q handled Model Context Protocol servers: it read an MCP configuration file from the open workspace and automatically launched the servers it defined. Because those servers run as local processes that inherit the developer's full environment, a single config file committed to a repo could reach AWS keys, cloud tokens, API secrets, and SSH agent sockets, turning a git clone into a full compromise. Amazon has patched the issue and published an advisory.
The crypto prediction market Polymarket says attackers stole close to $3 million from users after compromising a third-party vendor and injecting a malicious script into the platform's website. The script ran on the live site and prompted users connecting their wallets to approve transactions that drained their funds; researchers traced roughly $2.94 million taken from around a dozen accounts and bridged into Ethereum. Because the attack rode in through a trusted frontend dependency rather than Polymarket's own systems, it was invisible to users. Polymarket removed the dependency, contained the incident, and pledged full refunds. It was the platform's second security incident in two months.
Researchers at Island found that a popular Chrome extension, "Adblock for YouTube," with more than 10 million installs and a Featured badge, contains the machinery to run arbitrary JavaScript on any website the user visits. The extension works as advertised, but it can fetch a rule from its server that creates script elements with attacker-supplied content, giving access to page data, sessions, and forms. The capability is dormant, not absent: switching it on takes a single server-side change, with no extension update and no store review. The add-on changed ownership years ago, requests access to all sites, and is linked to other extensions previously pulled for malware.
Researchers at Novee disclosed Cordyceps, a systemic class of weaknesses in CI/CD pipelines, especially GitHub Actions workflows, that lets an attacker with nothing more than a free account hijack a project's build and release process. The danger is not a single bug but how workflows chain together: an untrusted pull request or comment feeds a low-privilege workflow whose output flows into a higher-privilege one, ending in stolen credentials, poisoned artifacts, or malicious releases. A scan of 30,000 repositories found over 300 fully exploitable, with fixes confirmed by Microsoft, Google, Apache, Cloudflare, and the Python Software Foundation. Standard scanners miss it because they check files in isolation.
Security firm AIR showed how easily AI agent skills can be weaponized by building a benign-looking design skill, publishing it to marketplaces, and promoting it with an Instagram ad until it reached roughly 26,000 agents, including some on corporate accounts. Every skill-scanning tool they tested, including offerings from Cisco and Nvidia, marked it safe. The trick is that the skill itself stays clean but tells the agent to fetch instructions from an external page the attacker controls, which passes review while pointing at harmless content and can be swapped for a malicious install script later. Skills load into an agent with the same authority as a user's prompt.
JFrog found malicious npm packages that impersonate PostCSS build tools to drop a multi-stage Windows remote-access trojan on developer machines. One package, postcss-minify-selector-parser, is named to look like the widely used postcss-selector-parser library, which sees over 127 million weekly downloads, and even lists the real package as a dependency to seem plausible during a quick review. Once installed, it writes and runs a PowerShell script that pulls down the trojan. A second cluster of five packages delivers a dropper during npm install, with one server-side component that only serves the payload to victims matching a specific signature. Affected developers should remove the packages and rotate credentials.
Attackers compromised the build pipeline of ShapedPlugin, a WordPress plugin maker, and slipped malware into legitimate updates delivered to paying customers through the vendor's own update system. The tainted releases install a fake plugin that impersonates WooCommerce components, steals site credentials, and gives attackers the ability to write files remotely. Three paid plugins are affected: Product Slider Pro for WooCommerce, Real Testimonials Pro, and Smart Post Show Pro. The backdoor was injected into Pro builds on May 21, with the first customer reports on June 10. Versions on WordPress.org stayed clean, pointing to a compromise of the vendor's release infrastructure rather than the plugins themselves.
Aikido Security uncovered a coordinated campaign of at least 15 malicious plugins on the JetBrains Marketplace that pose as AI coding assistants but secretly steal the AI provider API keys developers enter. The plugins offer real features like chat, code review, and commit messages, so they work as advertised, but the moment a user pastes in an OpenAI, DeepSeek, or SiliconFlow key and clicks Apply, the key is silently sent to an attacker server over plain HTTP, with no prompt. The campaign has run since late October 2025, with new plugins as recent as June 10, and uses inflated downloads and fake reviews. Separately, malicious Chrome extensions were found capturing chatbot conversations.
Attackers hijacked the npm account of a former contributor to Mastra, a popular open-source framework for building AI applications, and in an 88-minute automated burst republished 144 packages under the @mastra scope with a hidden malicious dependency. The poisoned dependency, a fake clone of a date library, runs at install time: it disables TLS checks, downloads a second-stage cryptocurrency-stealing trojan, runs it as a detached process, and deletes itself. Because @mastra/core alone sees over 900,000 weekly downloads and the payload fires on install, anyone who installed an affected version since June 16 could be compromised before importing anything. npm has pulled the malicious versions.
TrustStrike Labs