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 29 of 2026 · Retrospective · Covers 7 July 2026 11:00 to 14 July 2026 11:00 UTC

Read on the web

Top developments

Injective SDK on npm infected with cryptocurrency wallet stealer

BleepingComputer · Analysis · SCA

CISA orders feds to patch max severity ColdFusion flaw by Friday

BleepingComputer · Advisory · Remediation

Microsoft patches RoguePlanet Defender zero-day vulnerability

BleepingComputer · Advisory · Remediation

jscrambler npm package publishes malicious preinstall binary

StepSecurity · Analysis · SCA

Ubiquiti warns of new max severity UniFi OS vulnerability

BleepingComputer · Analysis · Remediation

SAST

Antiproof: Synthesizing Vulnerability Detectors and Proofs of Exploitability

arXiv · Research · SAST

Toward Inferring Accurate Context-free Grammars for Big Languages in a Black-box Setting

arXiv · Research · DAST, SAST

SeedSmith: LLM-Driven Seed Synthesis for Directed Fuzzing

arXiv · Research · Agentic SDLC, SAST

SCA

Skills That Don't Exist: A Large-Scale Study of Hallucinated Skill Recommendation in LLM Agents

arXiv · Research · SCA

Which Neurons Detect Malicious Code? A Probing Study of LLM Security Knowledge

arXiv · Research · SCA

Do These Violent Delights Have Violent Ends? Measuring the Post-Merge Fate of Agentic Code

arXiv · Research · Agentic SDLC, SCA

Researcher Details WhatsApp-to-Host Attack Chain Using Three OpenClaw Flaws

The Hacker News · Research · Remediation, SCA

DAST

Baselines Before Architecture: Evaluating Coding Agents for Autonomous Penetration Testing

arXiv · Research · Agentic SDLC, DAST

VEXAIoT: Autonomous IoT Vulnerability EXploitation using AI Agents

arXiv · Research · AI Security, DAST

ScopeJudge: Cost-Aware Pre-Execution Gating for Offensive Security Agents

arXiv · Research · DAST

Finding and Understanding Miscompilation Bugs in the Solidity Compiler

arXiv · Research · DAST

Agentic SDLC

'Ghostcommit' hides prompt injection in images to fool AI agents, steal secrets

BleepingComputer · Research · Agentic SDLC, AI Security

Trust but Verify? Uncovering the Security Debt of Autonomous Coding Agents

arXiv · Research · Agentic SDLC

AutoTrace: From Patches to Triggers via Agentic Interprocedural Exploration

arXiv · Research · Agentic SDLC, Remediation

Predicting Acceptance and Review Effort in Human and Agent Pull Requests

arXiv · Research · Agentic SDLC

Remediation

CISA orders feds to prioritize patching Langflow auth bypass flaw

BleepingComputer · Analysis · AI Security, Remediation

Zimbra urges customers to patch critical web client XSS flaw

BleepingComputer · Advisory · Remediation

Writer AI Flaw Could Let Agent Previews Leak Session Tokens Across Tenants

The Hacker News · Incident · Remediation

Patch for Windows Defender 0-day could allow attackers to fill hard disk

Ars Technica: Security · Analysis · Remediation

Trust

Policy-Conditioned Constrained Decoding for Column-Level Access Control in Text-to-SQL

arXiv · Research · Trust

Cardano's Voltaire Governance: Complete Specification and Research Program

arXiv · Research · Trust

Falsifiable Release Gates for Self-Improving Systems: Standing Invariants at Scale

arXiv · Research · Trust

The Replicant in Your Directory: AI Agents and the Identity Security Gap

BleepingComputer · Analysis · AI Security, Trust

Models

Breaking Refusal in the First Half: A Mechanistic Study of the Prefill Jailbreak

arXiv · Research · Models

Agent-Safety Evaluations as Load-Bearing Evidence: A Vendor-Neutral, Cross-Harness Reconstructability Metric

arXiv · Research · Models

Grok 4.5 Model Card

Cursor · Analysis · Models

Input-Aware Dynamic Backdoor Attack Against Quantum Neural Networks

arXiv · Research · Models

AI Security

Isolation as a First-Class Principle for LLM-Agent System Safety: Concepts, Taxonomy, Challenges and Future Directions

arXiv · Research · AI Security

Now, defenders are embracing the prompt injection, too

Ars Technica: Security · Analysis · AI Security

Mako: A Self-Evolving Agentic Operating System (SE-AOS) for Autonomous Web Exploitation

arXiv · Research · AI Security

SingGuard-NSFA: Extensible Guardrails for Agentic AI via Generative Reasoning and Real-Time Classification

arXiv · Research · AI Security

Advisories in this window

  • 503 npm advisories · 484 critical · 11 high · 6 moderate · 2 low
  • 51 Go advisories · 30 critical · 7 high · 12 moderate · 2 low
  • 49 PyPI advisories · 9 critical · 17 high · 22 moderate · 1 low
  • 47 Packagist advisories · 7 critical · 17 high · 21 moderate · 2 low
  • 15 Maven advisories · 2 critical · 7 high · 6 moderate
  • 13 RubyGems advisories · 3 high · 10 moderate
  • 9 Hex advisories · 5 high · 1 moderate · 3 low
  • 8 crates.io advisories · 1 critical · 2 high · 4 moderate · 1 low
  • 1 Multi-ecosystem advisory · 1 moderate

All advisories

Titles, links and short excerpts only; every item links to its source. Selection is automatic; the editor's note is AI-drafted and checked as described. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.