PREVIEW. This is how it will look in an inbox; it was not sent to you. Each real email carries its own unsubscribe link and Keryn's postal address. KERYN HEADLINES Weekly, week 28 of 2026 Retrospective Covers 30 June 2026 11:00 to 7 July 2026 11:00 UTC Web: https://keryn.dev/headlines/weekly/2026-W28/ TOP DEVELOPMENTS [1] CitrixBleed To Infinity And Beyond (Citrix NetScaler Pre-Auth Memory Overread CVE-2026-8451) watchTowr · Analysis · https://labs.watchtowr.com/citrixbleed-to-infinity-and-beyond-citrix-netscaler-pre-auth-memory-overread-cve-2026-8451 [2] JadePuffer ransomware used AI agent to automate entire attack BleepingComputer · Incident · https://bleepingcomputer.com/news/security/jadepuffer-ransomware-used-ai-agent-to-automate-entire-attack [3] Adobe patches seven max severity ColdFusion, Campaign flaws BleepingComputer · Advisory · https://bleepingcomputer.com/news/security/adobe-patches-seven-max-severity-coldfusion-campaign-flaws [4] BeyondTrust warns of critical flaws in remote access software BleepingComputer · Analysis · https://bleepingcomputer.com/news/security/beyondtrust-warns-of-critical-flaws-in-remote-access-software [5] Sysdig clocks first documented case of agentic ransomware CyberScoop · Incident · https://cyberscoop.com/sysdig-judepuffer-ai-agentic-ransomware-attack SAST [6] Multiple @immobiliarelabs Backstage Plugins Compromised on npm StepSecurity · Incident · https://stepsecurity.io/blog/immobiliarelabs-npm-packages-compromised [7] The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities arXiv · Research · https://arxiv.org/abs/2607.05743 [8] ShadowProbe: Language-Extensible Detection of Hidden Algorithmic Complexity Vulnerabilities arXiv · Research · https://arxiv.org/abs/2607.05474 [9] RustMizan: A Compilable, Contamination-Aware Benchmarking Framework for Rust Vulnerabilities arXiv · Research · https://arxiv.org/abs/2607.04729 SCA [10] 10 Layers Deep: How StepSecurity Stops TeamPCP's Trivy Supply Chain Attack on GitHub Actions StepSecurity · Incident · https://stepsecurity.io/blog/10-layers-deep-how-stepsecurity-stops-teampcps-trivy-supply-chain-attack-on-github-actions [11] Mass npm Supply Chain Attack: 20 Leo Platform Packages Compromised StepSecurity · Incident · https://stepsecurity.io/blog/mass-npm-supply-chain-attack-20-leo-platform-packages-compromised [12] Secure Registry now tells you which machine pulled a compromised package StepSecurity · Incident · https://stepsecurity.io/blog/secure-registry-now-tells-you-which-machine-pulled-a-compromised-package [13] 15 Malicious JetBrains Plugins Stole AI API Keys from 70,000 Developers StepSecurity · Incident · https://stepsecurity.io/blog/jetbrains-malicious-plugins-ai-api-key-theft DAST [14] Using OAI Overlay to Enhance REST API Fuzzing arXiv · Research · https://arxiv.org/abs/2607.04325 [15] Execution Divergence Graphs:Effective Discovery of Control-Flows from Execution Traces as Fuzzing Feedback arXiv · Research · https://arxiv.org/abs/2607.03396 [16] Large Language Models for Fuzz Testing in Microservices: A Systematic Literature Review arXiv · Research · https://arxiv.org/abs/2609.04219 [17] From Battlefield to Boardroom: Strategic Red Teaming as an Epistemic Governance Instrument in the Age of AI arXiv · Research · https://arxiv.org/abs/2607.01913 AGENTIC SDLC [18] SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review arXiv · Research · https://arxiv.org/abs/2607.06065 [19] Can Large Language Models Generate Observability-Aware Code? arXiv · Research · https://arxiv.org/abs/2607.05785 [20] Unicode TAG-Block Concealment of Tool-Metadata Payloads in the Model Context Protocol: An Approval-View Fidelity Gap Across Three Independent Server Implementations arXiv · Research · https://arxiv.org/abs/2607.05744 [21] SkillCloak Lets Malicious AI Agent Skills Evade Static Scanners with Self-Extracting Packing The Hacker News · Research · https://thehackernews.com/2026/07/new-skillcloak-technique-lets-malicious.html REMEDIATION [22] Beyond Refusal: A Same-Lineage Study of Aligned and Abliterated LLMs for Vulnerability Analysis arXiv · Research · https://arxiv.org/abs/2607.05842 [23] Opera GX Flaw Let Malicious Sites Auto-Install Mods to Steal Data From Visited Pages The Hacker News · Research · https://thehackernews.com/2026/07/opera-gx-flaw-let-malicious-sites-auto.html [24] Cisco finally confirms attackers exploiting Unified CM flaw BleepingComputer · Analysis · https://bleepingcomputer.com/news/security/cisco-finally-confirms-attackers-exploiting-unified-cm-flaw [25] GPT-5.5-Cyber built a zlib fuzzing lab in a day Trail of Bits · Analysis · https://blog.trailofbits.com/2026/07/02/field-reports-from-patch-the-planet TRUST [26] Cool New Whitepaper SPDX · Research · https://spdx.dev/cool-new-whitepaper [27] Governed Individuation: Cryptographically Decoupling an Agent's Learning from Its Authority arXiv · Research · https://arxiv.org/abs/2607.04613 [28] Agentao: A Policy-Governed Runtime Harness for Embeddable Tool-Using LLM Agents arXiv · Research · https://arxiv.org/abs/2608.13574 [29] Macro-Prudential AI Governance: A Two-Layer Early Warning and Response System for Frontier AI arXiv · Research · https://arxiv.org/abs/2607.03542 MODELS [30] Auditing of Unlearning Algorithms arXiv · Research · https://arxiv.org/abs/2607.05898 [31] Two Sides of the Same Coin: Learning the Backdoor to Remove the Backdoor arXiv · Research · https://arxiv.org/abs/2607.05748 [32] Statistical Adversaries: Natural Backdoor-like Adversarial Features in Clean Vision Datasets arXiv · Research · https://arxiv.org/abs/2607.05516 [33] Privacy-Preserving Robustness Verification for Neural Networks arXiv · Research · https://arxiv.org/abs/2607.05251 AI SECURITY [34] The ‘first’ AI-run ransomware attack still needed a human TechCrunch: Security · Incident · https://techcrunch.com/2026/07/06/the-first-ai-run-ransomware-attack-still-needed-a-human [35] Phantom Squatting: AI-Hallucinated Domains as a Software Supply Chain Vector Palo Alto Networks · Analysis · https://unit42.paloaltonetworks.com/phantom-squatting-hallucinated-web-domains [36] Microsoft Warns Poisoned MCP Tool Descriptions Can Make AI Agents Leak Data The Hacker News · Incident · https://thehackernews.com/2026/06/microsoft-warns-poisoned-mcp-tool.html [37] When Agents Remember Too Much: Memory Poisoning Attacks on Large Language Model Agents arXiv · Research · https://arxiv.org/abs/2607.06595 ADVISORIES IN THIS WINDOW 286 npm advisories · 215 critical · 40 high · 30 moderate · 1 low 81 Go advisories · 14 critical · 33 high · 29 moderate · 5 low 61 crates.io advisories · 2 critical · 9 high · 39 moderate · 11 low 55 Maven advisories · 11 critical · 29 high · 11 moderate · 4 low 51 PyPI advisories · 5 critical · 32 high · 5 moderate · 9 low 36 Packagist advisories · 3 critical · 15 high · 14 moderate · 4 low 10 NuGet advisories · 5 high · 4 moderate · 1 low 5 Hex advisories · 1 critical · 4 moderate 1 RubyGems advisory · 1 high All advisories: https://keryn.dev/headlines/advisories/ Titles, links and short excerpts only; every item links to its source. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.