Search: github.com

A sophisticated software supply chain attack named ChainDrop has infected hundreds of npm packages, including popular caching libraries with millions of weekly downloads. Beginning August 4, 2026, attackers compromised a GitHub account of a keyv package maintainer, injecting malicious code into legitimate repositories. The malware executes credential-stealing payloads targeting developer workstations and CI/CD runners, harvesting npm tokens, GitHub credentials, cloud access keys, SSH keys, and database credentials. Using stolen credentials, the worm self-propagates by compromising additional repositories and publishing poisoned packages with valid provenance attestations. ChainDrop employs Bun runtime for execution, establishes persistence through developer tool configurations, and exfiltrates encrypted data using blockchain-based command-and-control infrastructure. This campaign represents an evolution of the Shai-Hulud npm worm.

The npm package agenthub-multiagent-mcp version 1.57.0 contains malicious code that establishes a persistent WebSocket connection to a hardcoded remote server. This connection allows the remote server to execute arbitrary commands with disabled permission checks on the user's system, including file operations and shell commands. The package also installs OS-level persistence mechanisms on Windows, macOS, and Linux to maintain remote access across reboots. The package falsely claims to be associated with a legitimate GitHub repository, while the actual control domain and author differ.

A 710 KB JavaScript payload was discovered in the compromised keyv@6.0.0 package, representing a newer variant of Shai-Hulud with enhanced obfuscation techniques. The malicious code operates with four primary objectives: harvesting credentials from local systems, CI environments, cloud platforms, Kubernetes, and Vault; exfiltrating encrypted data through dynamic HTTPS endpoints or public GitHub repositories; leveraging stolen npm credentials to publish infected patch releases across accessible packages; and exploiting GitHub credentials with GitHub Actions to compromise repositories and extract additional credentials. The campaign demonstrates sophisticated supply chain attack capabilities, targeting the npm ecosystem and development infrastructure. Multiple components were identified including obfuscated JavaScript files, VS Code configuration files, and injected GitHub Actions workflows, indicating a comprehensive approach to credential theft and lateral movement across development environments.

A sophisticated supply-chain attack campaign named Shai-Hulud has compromised over 400 npm packages across 1700+ versions, beginning with keyv and cacheable libraries. The malware operates as a self-propagating worm that collects credentials from local filesystems, CI/CD environments, cloud platforms, Kubernetes clusters, and HashiCorp Vault. It exfiltrates stolen data through dynamic HTTPS endpoints or public GitHub repositories, then uses compromised npm tokens to publish infected versions of all writable packages. The campaign also injects execution hooks into GitHub repositories via VS Code and Claude configuration files, harvests GitHub Actions secrets through injected workflows, and includes a targeted attack against npm trusted publishing flows. Command and control infrastructure leverages Ethereum smart contracts and GitHub commit messages for resilience.

A vulnerability was determined in OpenHands up to 0.62.0. The affected element is the function initialize_repo of the file OpenHands/resolver/send_pull_request.py. This manipulation causes command injection. Remote exploitation of the attack is possible. The vendor deleted the original GitHub issue report. It appears that the affected path/file got removed in version 1.7.0.

On August 4, 2026, a sophisticated supply chain attack compromised the keyv npm package maintainer, deploying CHAINDROP, a self-propagating worm that automatically backdoors packages using stolen npm credentials. Over 400 npm packages were infected, affecting more than 1.3 billion monthly downloads. The worm executes via preinstall hooks, deploys across Linux, macOS, and Windows platforms, and harvests credentials from over 300 patterns targeting AI tooling, cloud providers, GitHub tokens, and npm credentials. CHAINDROP uses Ethereum smart contracts for C2 resolution and propagates by publishing trojanized versions of packages the compromised maintainer can access. The payload is heavily obfuscated and contains Dune-themed references consistent with previous Shai-Hulud campaigns.

PraisonAI is a multi-agent teams system. In versions prior to 4.6.40, the bundled Claude GitHub Actions workflow is vulnerable to command injection because it embeds an attacker-controlled pull request branch name into a Bash run: block without quoting or validation. Additionally, the workflow allows any @claude comment to trigger the job regardless of whether the commenter is a trusted collaborator. An outside contributor can open a pull request from a fork whose branch name contains shell metacharacters and comment @claude, causing Bash to execute arbitrary shell code in the GitHub Actions runner. Because these commands run in a job holding a GitHub App token with write permissions, OIDC access, and gh/git access, the injection can be chained through $GITHUB_PATH to compromise later privileged steps, enabling repository writes, pull request and issue manipulation, or OIDC-token abuse. This issue has been fixed in version 4.6.40.

When you learn that a compromised package executed on one of your build hosts, muscle memory takes over: revoke the npm token, rotate the GitHub PAT, cycle the cloud keys. That reflex has been correct in almost every supply-chain incident I have worked. In the keyv/cacheable compromise that has been unfolding since yesterday, it is the one thing you should not do first — because revoking the stolen token is exactly what arms the payload.

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (409d11825a61bed068425296e83dcc0c87427579f7b5d724ec52e5e87465f2b5) The package ships a prebuilt darwin/x64 Mach-O binary containing a Go agent (module github.com/ccfly/rescue) that, when executed, connects to a hardcoded remote broker at cc.hn / ccfly over WebSocket (ws://ccfly, wss://, https://cc.hn) using github.com/gorilla/websocket, spawns a pseudo-terminal via github.com/creack/pty, and pipes the WebSocket stream into that PTY. Symbols include agent.serve, agent.startPTY, agent.waitAuthorized, agent.wsBase and unixPTY.Read/Write/Resize/Close, giving the remote party interactive shell control of the host. A second stage (setup.downloadCcfly, setup.fetchAndExtract, setup.npmTarballURL, setup.extractFromTarGz, setup.runCcflyInstall, setup.redeemEnrollToken, setup.(*brokerClient).escalateToRescue, with a rodata reference to https://registry.npmjs.org) fetches and executes a further ccfly npm tarball after the broker approves an enrollment token — the additional code executed on the host is chosen by the remote broker, not the installer. The binary also inspects shell RC files and /etc/hosts (setup.scanShellProfilesForProxy, setup.checkProxyResidue, setup.checkHostsResidue, setup.runEnvChecks, setup.envReport) and POSTs an environment report back to the broker (setup.(*brokerClient).post; rodata latest.zshrc, export.ccfly).

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (9434f2e4ac17da4d2b9dff0c36ed8b3c5cc3f76882b9898b768e78ba982bbac2) The package declares a preinstall lifecycle hook that runs index.js on `npm install`. index.js collects host identifiers and system files (os.hostname(), os.userInfo(), home directory, DNS servers, /etc/passwd, /etc/hosts, and package.json contents) and POSTs them via HTTPS to the hardcoded subdomain lncqub6t1loht8b887c8hvcjva13ptdi.oastify.com — a Burp Collaborator out-of-band interaction host used for reconnaissance. There is no advertised functional purpose in the package that would justify this data collection; the code path exists solely to transmit installer host and credential-adjacent file contents to an attacker-controlled endpoint. ## Source: ghsa-malware (87310aa13634f4a26c92164452e9ff0c95d3ee5924feb2f6b5c48354d32ee482) Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it.

An active supply chain attack has compromised the keyv and cacheable npm packages, affecting tens of millions of weekly downloads. The attack began on August 4, 2026, when the maintainer account Jaredwray was compromised, enabling attackers to publish malicious code across multiple packages. The malware deploys through a preinstall hook that downloads a Bun runtime and executes obfuscated payloads designed to harvest cloud credentials from AWS, GCP, Azure, HashiCorp Vault, Kubernetes, GitHub Actions, and npm tokens. The threat exhibits worm-like behavior by using stolen npm tokens to republish trojanized versions of additional packages beyond the original namespaces. Stolen credentials are exfiltrated to attacker-controlled GitHub repositories via DNS-resolved destinations, with persistence mechanisms planted in developer environments through .claude and .vscode hooks.

Attackers compromised a GitHub maintainer account controlling keyv, cacheable, flat-cache, and file-entry-cache Node.js packages that collectively receive over a billion downloads monthly. Malicious code was pushed directly to the main branch and automatically published to npm with valid signatures. A hidden preinstall script downloads a Bun runtime to execute an obfuscated payload that harvests npm, GitHub, AWS, Kubernetes, and Vault credentials, scans for SSH keys and environment files, and exfiltrates data to attacker-controlled GitHub repositories and Ethereum smart contracts. The worm then uses stolen npm tokens to infect additional packages autonomously. This self-propagating attack, tracked as ChainDrop, belongs to the Shai Hulud family responsible for previous campaigns targeting TanStack, Mistral AI, and OpenSearch packages in May 2026.

** UNSUPPORTED WHEN ASSIGNED ** Stack-based Buffer Overflow vulnerability in Apache Lucy. This issue affects Apache Lucy: all versions. As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative or restrict access to the instance to trusted users. Lucy is now maintained outside of the ASF at https://github.com/lucysearch . This issue has been fixed in 0.8.0 there. NOTE: This vulnerability only affects products that are no longer supported by the maintainer.

** UNSUPPORTED WHEN ASSIGNED ** Deserialization of Untrusted Data vulnerability in Apache Lucy. This issue affects Apache Lucy: all versions. As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative or restrict access to the instance to trusted users. Lucy is now maintained outside of the ASF at https://github.com/lucysearch . 0.8.0 is no longer affected by this issue, because the offending feature has been removed there. NOTE: This vulnerability only affects products that are no longer supported by the maintainer.

** UNSUPPORTED WHEN ASSIGNED ** Uncontrolled Recursion vulnerability in Apache Lucy. This issue affects Apache Lucy: all versions. As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative or restrict access to the instance to trusted users. NOTE: This vulnerability only affects products that are no longer supported by the ASF. Lucy is now maintained outside of the ASF at https://github.com/lucysearch

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (fc34f69b5a68c549feccfb32cc2dbfbf2d4c787a4876241027b4299305fd2c53) The package is a platform-specific leaf (linux/arm64) whose only shipped artifact is a 12 MB stripped Go binary at bin/sxt. The package.json omits repository, author, main, bin, files, and scripts fields, and its license field points at github.com/ddos798/claude_control — a self-labeled 'claude_control' project. Strings in bin/sxt reveal a full WebRTC stack (pion/webrtc/v4, pion/turn/v5, pion/ice/v4, coder/websocket, go-qrcode) plus a hardcoded signaling/relay endpoint https://relay.sextant.top/install, giving the binary the shape of a remote-control agent that pairs the installer's host to an author-controlled peer over WebRTC data channels. Adjacent strings 'sk-ant-', 'CLAUDE_CONFIG_DIR', 'https://api.anthropic.com/v1/models', 'https://claude.ai/install.sh', and '@google/gemini-cli' show the binary is positioned to read Anthropic API keys and Claude Code / Gemini CLI configuration from the installer and hand them to the remote peer through the established channel. An additional string http://ip-api.com/json/ provides public-IP geolocation of the host for the remote operator. The combination of an opaque metadata-only npm manifest, a bundled native binary embedding a full peer-to-peer control stack, a hardcoded author relay, and embedded references to developer AI-credential material is a backdoor with credential-theft capability against the installer.

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (9d2e5ab6503eb2b9c12bf64e34cb2fe12227c07b179231ebe23f13b8866181f1) The tarball contains a single 13.4 MB Mach-O darwin/amd64 Go binary at bin/sxt and a package.json with no source, no main, no bin mapping, and no scripts — the package is a platform-artifact shipped for consumption as an optionalDependency by a parent sextant-cli. Strings extracted from bin/sxt show it embeds github.com/coder/websocket and github.com/creack/pty together with the hardcoded relay URL wss://relay.sextant.top; that network-to-PTY dataflow is a remote-shell backdoor driven by an attacker-controlled WebSocket relay on the installer's host. The same binary contains the Anthropic API-key regex sk-ant-[a-z0-9]+-[A-Za-z0-9_-]{40,} alongside CLAUDE_CONFIG_DIR=, https://claude.ai/, and https://api.anthropic.com/v1/models, indicating the binary reads the installer's Claude configuration directory and fingerprints sk-ant-* credentials for exfiltration through the same relay. Additional strings embed http://ip-api.com/json/?fields=status,message,country,countryCode,city,timezone,isp,org,as,proxy,hosting,mobile,query, profiling the installer's public IP, city, ISP, org, ASN, and hosting/proxy flags. package.json declares its license as SEE LICENSE IN https://github.com/ddos798/claude_control, and the binary references https://claude.ai/install.sh and https://registry.npmjs.org/sextant-cli/latest for self-update. Delivering behavior only as a stripped compiled binary conceals the network-to-PTY, credential-regex, and geolocation logic from JS-level inspection.

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (aab426dad8f1e5722f2836debe2f93730df98b7c109865426208edb687955b30) sextant-cli-linux-amd64 ships a Linux amd64 Go binary that, on execution, connects to a hardcoded WebRTC signaling relay at wss://relay.sextant.top and exposes a local interactive shell to a remote controller. The binary uses creack/pty and exec.Command to spawn /bin/bash, /bin/sh, or /bin/zsh under a PTY and forwards stdio through WebRTC data channels, giving a remote party on the sextant.top relay a full interactive shell on the installer's host. The binary also contacts http://ip-api.com/json/ over cleartext HTTP to collect the host's public IP and geolocation and reports it back through the same control channel. Embedded strings referencing /etc/systemd and 'sextantsystemdbootoutdefault.claudekey' indicate a subcommand that installs the agent as a systemd unit for boot persistence. package.json's license URL points to https://github.com/ddos798/claude_control, and the binary internally references 'claude_control' and api.anthropic.com, consistent with a tool designed to remotely pilot the installer's host and Claude Code sessions. Regardless of the tool's advertised purpose, a network-sourced controller driving a PTY on the installer's machine, with systemd persistence support and host reconnaissance to a third-party geolocation service, constitutes a persistent remote backdoor into the installer host.

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (23978b08a7e3f8af97d2c8060e7fc3aca9a18d9f3435163a9d3ab97ac0b9d4d0) Package ships a darwin/arm64 Go binary that opens an outbound WebSocket to a hardcoded author-controlled relay at wss://relay.sextant.top and links a PTY library (github.com/creack/pty) with a WebSocket client (github.com/coder/websocket), implementing the network-driven pseudo-terminal remote-shell pattern that yields full-host code execution to the operator of that relay. The binary additionally binds listeners on 0.0.0.0:7280, 0.0.0.0:7281, and 0.0.0.0:9000 (a 'ws://%s:%s%s' format string is present), exposing the PTY/WebSocket surface to the local network rather than loopback. Embedded strings include a regex specifically matching Anthropic API keys ('sk-ant-[a-z0-9]+-[A-Za-z0-9_-]{40,}') and references to CLAUDE_CONFIG_DIR and ~/.claude paths, allowing extraction of the installer's Anthropic credentials, which are then reachable via the same relay channel. A separate POST to https://relay.sextant.top/install combined with a call to http://ip-api.com/json/?fields=status,message,country,countryCode,city,timezone,query reports installer geolocation and host identity to the author. The npm-facing metadata describes the package only as 'Sextant (sxt) binary for darwin/arm64', while the license field points to a GitHub repository named 'claude_control' — the shipped functionality (remote control of Claude Code / Gemini CLI sessions plus API-key exfiltration) is not disclosed in the package description or README.

In this article Attack chain overview Mitigation and protection guidance Indicators of compromise (IOC) Microsoft Defender XDR detections Advanced hunting queries Learn more Microsoft Threat Intelligence identified a large-scale npm supply chain attack affecting more than 400 packages across multiple unrelated publishers, including packages associated with major enterprise software ecosystems such as keyv, flat-cache, cache-manager, and others. The malicious releases contain a Mini Shai-Hulud variant, a self-propagating credential-stealing worm delivered through a large, heavily obfuscated Bun-based JavaScript payload. The malware typically executes automatically through an npm preinstall lifecycle hook before package installation completes. Once executed, the malware searches developer workstations and continuous integration and continuous delivery (CI/CD) environments for npm, GitHub, cloud, and infrastructure credentials. It uses recovered identities to authenticate to npm, GitHub, Amazon Web Services (AWS), Kubernetes, and HashiCorp Vault, enabling it to enumerate packages, repositories, workflow secrets, cloud parameters, and secret-store values. Collected data is encrypted and transmitted through an attacker-controlled HTTPS endpoint, with GitHub repositories serving as a fallback exfiltration channel. The payload’s most significant capability is automated propagation. After obtaining an npm publishing token, it enumerates packages available to the compromised identity, downloads their latest tarballs, inserts the malware and setup loader, adds a preinstall hook, increments the patch version, and republishes the modified packages. The malware can also use stolen GitHub credentials to inject Claude and Visual Studio Code configuration files into repositories, establishing persistence and creating an additional developer-to-developer infection path. In this blog, we’re sharing our analysis of this supply chain attack, along with protection, detection, amd hunting guidance. Organizations that installed an affected package with lifecycle scripts enabled should treat the associated developer workstation or build runner as potentially compromised. Investigations should prioritize credentials accessible to the affected identity, unauthorized npm releases, unexpected repository or workflow modifications, suspicious cloud and secret-store access, and artifacts produced by affected build systems. Organizations should revoke and rotate exposed credentials from a known-clean environment and rebuild affected systems and downstream artifacts from trusted sources. Attack chain overview The campaign appeared as a rapid sequence of unauthorized patch releases across more than 400 npm packages maintained by otherwise unrelated publishers. Many malicious versions had no corresponding source-code commit, pull request, tag, or legitimate release, indicating that the attackers modified and published package tarballs directly rather than compromising each public source repository. Affected releases typically added a preinstall lifecycle script that launched a malicious file, setup.mjs, contained within the package, which launched the large, obfuscated Bun JavaScript bundle included in the package. Because npm runs preinstall scripts before installation completes, the payload could execute on developer workstations and build runners before application tests or conventional security checks began. After execution, the malware performs the following actions: Determines whether it is running on a developer workstation or in a CI/CD environment. On workstations, it detaches itself to continue after installation; on CI/CD systems, it remains in the active job to access workflow secrets, runner credentials, and OpenID Connect (OIDC) publishing permissions. Both paths could support further package or repository propagation when suitable credentials are found. Collects credentials from local files, environment variables, command-line tools, and GitHub A…