← All briefings

Daily Security Intelligence Briefing — August 9, 2026

Overall Daily Threat Level: HIGH.

Tier 1 — Priority Targets (Must-Read)

1. Shai-Hulud npm Supply Chain Worm Compromises 1,280+ Packages via GitHub Actions Pipelines

Source & Article Link: HackRead / TLDR InfoSec

Technical Severity / Domain: Software Supply Chain Security / Critical (Self-Propagating Worm)

Why You Must Read This: A new Shai-Hulud worm variant compromised over 1,280 npm packages (including Keyv) by injecting malicious commits into GitHub Actions build pipelines. The worm uses a preinstall hook to download Bun and execute an obfuscated loader that aggressively harvests AWS credentials, Kubernetes tokens, HashiCorp Vault secrets, npm tokens, and SSH keys from developer machines and CI runners.

Recommended Immediate Action: Rotate all developer and cloud tokens immediately if any compromised dependency versions were pulled, and perform manual log reviews rather than relying solely on scheduled vulnerability scans.

2. Over 4,400 Rockwell PLCs Exposed Online; 22 Found in Targeted Water Utility Cities

Source & Article Link: The Hacker News / TLDR InfoSec

Technical Severity / Domain: Industrial Control Systems (ICS/SCADA) / Critical Infrastructure Risk

Why You Must Read This: Forescout identified over 4,400 internet-facing Rockwell Automation PLCs globally, including 22 located in US water utility cities recently subjected to cyberattacks. Attackers compromised controllers via open Ethernet/IP port 44818 without exploiting complex zero-days, locking operators out by altering logic configurations.

Recommended Immediate Action: Immediately disconnect all industrial controllers from direct public internet exposure behind perimeter firewalls or industrial VPNs, and apply Rockwell Advisory SD1790 to clear unauthorized locks.

3. Pre-Auth Stack Buffer Overflow on Mercusys MB115-4G Router (CVE-2026-12495)

Source & Article Link: Rotcee Hardware Research / TLDR InfoSec

Technical Severity / Domain: Hardware & Firmware Security / Critical (CVE-2026-12495)

Why You Must Read This: Hardware reverse engineering of Amazon Spain's best-selling router identified a pre-authentication vulnerability in /usr/bin/httpd. The http_gdpr_decrypt function copies up to 2,048 bytes of AES-decrypted, attacker-controlled data into a 512-byte caller stack buffer with no bounds check. Static analysis in Ghidra and Ghidriff confirmed the flaw persisted unpatched across firmware versions v1.7.0 through v1.9.0.

Recommended Immediate Action: Upgrade Mercusys MB115-4G router units to firmware v1.11.0 or newer, which adds bounds checking, HMAC verification, and transitions login decryption from RSA to ECC.

4. Greatness Phishing Service Spoofs RingCentral to Steal M365 Session Tokens

Source & Article Link: TLDR InfoSec — Greatness RingCentral Phish

Technical Severity / Domain: Cloud IAM & Phishing / High

Why You Must Read This: The Greatness phishing platform bypassed email gateway filters by spoofing RingCentral voicemail alerts to target organizations that had blanket-whitelisted the domain, despite failing SPF, DKIM, and DMARC checks. Once clicked, victims were routed through Adversary-in-the-Middle (AiTM) proxies to steal MFA tokens, enabling stealth data exfiltration across Microsoft 365.

Recommended Immediate Action: Replace broad safe-sender exclusions with strict authentication enforcement, hunt for anomalous MFA sign-ins originating from hosting providers, and revoke active user tokens upon suspected exposure.

Tier 2 — Essential Intelligence (Summaries & Analysis)

1. Cloudflare Open-Sources pvcli Privacy Proxy Debugger

Domain: Cloud Security & Privacy Engineering

Subject / Topic Summary: Cloudflare open-sourced pvcli, a CLI utility designed to simplify debugging Oblivious HTTP (OHTTP) privacy proxies by automating binary encoding, encryption, and multi-party request routing in a single command.

Key Technical Takeaways:

Replaces manual hex-encoded key parsing with verbose step-by-step logging across client, relay, and gateway hops.

Supports mTLS to relays and custom relay headers, with a roadmap including MASQUE transports and post-quantum OHTTP.

Essential Direct Quote: "pvcli simplifies debugging privacy-preserving protocols like Oblivious HTTP by handling the binary encoding, encryption, and multi-party request routing in a single curl-style command."

Source: Cloudflare Engineering Blog

2. Agentic Incident Response Notebooks via marimo Framework

Domain: Incident Response & SOC Automation

Subject / Topic Summary: The DNB Cyber Defense Center built modular incident response playbooks using marimo notebooks and a custom SDK for cross-platform query translation across Splunk, Microsoft Defender, and SQL databases.

Key Technical Takeaways:

Leverages marimo's cell dependencies and caching capabilities for reproducible investigations.

Exposes the marimo-pair interface, allowing AI agents and human responders to collaboratively execute identical notebook workflows.

Essential Direct Quote: "The marimo-pair agentic interface allows agents to create and execute the same incident response notebooks as human responders."

Source: Cauchy Research / TLDR InfoSec

3. 77 Malicious Open VSX Extensions Harvest Developer Workstation Telemetry

Domain: IDE Security & Developer Tooling

Subject / Topic Summary: Manifold Security uncovered an "evil twin" campaign on the Open VSX registry involving 77 malicious extensions impersonating legitimate packages to harvest developer environment data.

Key Technical Takeaways:

Exfiltrated OS usernames, workspace file paths, Git commit hashes, and CI/CD details to mangorbit[.]com.

Used DNS TXT lookups for resilient command-and-control communication.

Essential Direct Quote: "The most aggressive variants actively exfiltrated OS usernames, workspace paths, Git commit hashes, and CI/CD infrastructure details to mangorbit[.]com, utilizing DNS TXT lookups for C2 resilience."

Source: TLDR InfoSec — Open VSX Campaign

4. Google DeepMind C-Suite Leadership Restructuring

Domain: Security Leadership & AI Governance

Subject / Topic Summary: Google DeepMind CEO Demis Hassabis transitioned to Chair of Google DeepMind and Chief Scientist of Alphabet, while Chief Scientist Jeff Dean departed after 27 years to launch Discovery Loop.

Key Technical Takeaways:

Signals consolidation of Alphabet's core AI research leadership under Hassabis amidst intensifying competitive pressure.

Essential Direct Quote: "Demis Hassabis moved to the role of chair of Google DeepMind and chief scientist of Alphabet, while Jeff Dean departed after 27 years to launch Discovery Loop."

Source: TLDR AI / Implicator

5. Cloudflare Agent Access Model (AAM) Reference Framework

Domain: Cloud Security & AI Identity Architecture

Subject / Topic Summary: Cloudflare published the Agent Access Model (AAM), redefining access control for AI software agents around task-specific ephemeral credentials and stateless policy enforcement.

Key Technical Takeaways:

Core principles include short-lived tokens, harness/network-level enforcement, evidence-based grant reviews, and unidirectional capability changes.

Essential Direct Quote: "By removing implicit trust in task execution graphs and evaluating every action against the task's state, AAM minimizes agent capabilities, thus reducing risk."

Source: Cloudflare Engineering Blog / TLDR AI

Tier 3 — Industry Noise & Awareness Only

• AgentHound Open-Source Tool: Offensive framework for fingerprinting and looting AI agent infrastructure (GitHub AgentHound).

• Beignet macOS Shellcode Utility: Converts .dylib files into position-independent macOS shellcode (GitHub Beignet).

• EndlessDoors Zbtlink Router Backdoor: Unpatchable C2 backdoor found in 100,000+ active routers (Decipher).

• Swiss Government SharePoint Breach: 200 accounts compromised in Microsoft SharePoint vulnerability exploit (TLDR InfoSec).

• Vanta Stealer Python Malware: Information stealer targeting browser credentials and Discord tokens (Point Wild Research).

• Pass-Ta-Key Google Passkey Attack: Unit 42 research on TPM-equipped Windows passkey extraction (TLDR InfoSec).

• English National Ballet CRM Breach: Customer relationship portal compromised, exposing contact information (BBC News).

• Anthropic $10B Cloud Capacity Deal: Anthropic contracted 6 years of cloud compute with Volta in Norway (TechCrunch).