<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Keryn Headlines: Agentic SDLC</title><link>https://keryn.dev/headlines/tag/agentic-sdlc/</link><description>AI coding agents in the development lifecycle and the risks they bring.</description><language>en</language><lastBuildDate>Sun, 22 Nov 2026 15:38:33 GMT</lastBuildDate><atom:link href="https://keryn.dev/headlines/tag/agentic-sdlc/feed.xml" rel="self" type="application/rss+xml"/><item><title>Forrester Recognizes Checkmarx in The Forrester Wave™: Agentic Development Security Tools, Q4 2026</title><link>https://checkmarx.com/blog/checkmarx-named-a-leader-in-the-forrester-wave-static-application-security-testing-solutions-q3-2025-2</link><guid isPermaLink="false">keryn-headlines-event-ev_65375520c1ea</guid><pubDate>Fri, 20 Nov 2026 17:03:20 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>Checkmarx scores highest in Current Offering category across ten evaluated vendors.</description></item><item><title>Replication of Agentic Security Events: Audit and Investigation (ROGUE-AI)</title><link>https://arxiv.org/abs/2611.26553</link><guid isPermaLink="false">keryn-headlines-event-ev_50f3355f9fc4</guid><pubDate>Thu, 19 Nov 2026 18:23:35 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Agents based on large language models (LLMs) can now delete files and send payments on behalf of their users. News media reports routinely describe failures of these agents (e.g., Replit&#39;s coding agent deleting a production database), but a report often only says what went…</description></item><item><title>Characterizing and Detecting Unfaithful Work Reports by Coding Agents</title><link>https://arxiv.org/abs/2611.25760</link><guid isPermaLink="false">keryn-headlines-event-ev_85941ac36402</guid><pubDate>Thu, 19 Nov 2026 09:01:02 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Coding agents increasingly carry out long-horizon development tasks, autonomously planning and executing multi-step actions that span files, tools, and sessions. Because developers rarely follow each action, and agents resuming the work depend on records of what was done, the…</description></item><item><title>Hardening JavaScript Programs against LLM-Based Vulnerability Detection</title><link>https://arxiv.org/abs/2611.25718</link><guid isPermaLink="false">keryn-headlines-event-ev_44aa4406e8f2</guid><pubDate>Thu, 19 Nov 2026 08:31:31 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>JavaScript programs are often distributed as source code, so attackers can analyze them for vulnerabilities. LLM-based vulnerability detection agents now make this analysis much easier. To mitigate this emerging threat, we present AntiAgent, a defense that lets developers…</description></item><item><title>Why the Harness Matters: Early Lessons From the Evo Code Scanning Agent Preview</title><link>https://snyk.io/blog/why-harness-matters-evo-code-scanning-agent</link><guid isPermaLink="false">keryn-headlines-event-ev_b7e22ed796aa</guid><pubDate>Thu, 19 Nov 2026 04:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>SAST</category><description>In early tests, Evo Code Scanning Agent’s security harness found more known vulnerabilities than Claude Security and Codex Security at the same compute budget, with lower precision. See what the results reveal about the role of the harness in AI code scanning.</description></item><item><title>Explore the Codex Security Plugin</title><link>https://openai.com/business/learn/codex-security-plugin-demo</link><guid isPermaLink="false">keryn-headlines-event-ev_168bef72de7b</guid><pubDate>Thu, 19 Nov 2026 00:00:00 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><description>Watch a demo of the Codex Security Plugin and see the workflow in action, from getting started to reviewing the results shown in the recording.</description></item><item><title>Your coding agents have a sandbox</title><link>https://endorlabs.com/learn/introducing-sandbox-posture-in-coding-agent-governance</link><guid isPermaLink="false">keryn-headlines-event-ev_b92be56a59dd</guid><pubDate>Wed, 18 Nov 2026 19:49:26 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>Your coding agents have a sandbox. Are your developers using it? Coding Agent Governance now reports whether the sandbox was on for every session.</description></item><item><title>Inside the Codex sandbox</title><link>https://openai.com/index/inside-the-codex-sandbox</link><guid isPermaLink="false">keryn-headlines-event-ev_5bd7761a1ab8</guid><pubDate>Wed, 18 Nov 2026 08:00:00 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><description>See how Codex runs commands on your machine while enforcing the filesystem and network limits you control across macOS, Linux, and Windows.</description></item><item><title>External sandboxes for coding agents</title><link>https://simonwillison.net/2026/Nov/17/external-sandboxes-for-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_e462d5b8dfb8</guid><pubDate>Tue, 17 Nov 2026 22:52:29 GMT</pubDate><category>Release</category><category>Agentic SDLC</category><category>AI Security</category><description>External sandboxes for coding agents I&#39;ve been exploring new ways to sandbox code that is executed by agents, with a focus on making it safer to try out local models without worrying that a mistake or prompt injection attack might cause them to take damaging actions on a…</description></item><item><title>The risks of device code authentication for AI coding agents</title><link>https://microsoft.com/en-us/security/blog/2026/11/17/the-risks-of-device-code-authentication-for-ai-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_be12f05d70ed</guid><pubDate>Tue, 17 Nov 2026 15:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Lab research shows how device code phishing, a technique that applies to any device-flow client, can extend to AI coding agents, using GitHub Copilot CLI as an illustrative example.</description></item><item><title>Securing and Auditing Long-Horizon LLM Agents: State-Transfer Auditing and Controls in the Agent Harness</title><link>https://arxiv.org/abs/2611.23346</link><guid isPermaLink="false">keryn-headlines-event-ev_a43b99fdd65c</guid><pubDate>Tue, 17 Nov 2026 10:36:04 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Long-horizon LLM agents across codebases, web browsers, and enterprise infrastructure rely on an execution harness to move state between the working context, session history, persistent memory, and external environment. While conventional execution traces log model prompts and…</description></item><item><title>Truthgate: Checking Coding Agents Against a Record They Cannot Edit - Cheating, Deterrence and Disclosure on Impossible Tasks</title><link>https://arxiv.org/abs/2611.23335</link><guid isPermaLink="false">keryn-headlines-event-ev_47dd7b7c9540</guid><pubDate>Tue, 17 Nov 2026 01:43:45 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>When a coding agent says it is done, is that true, and how would anyone know? We study this question in five steps; every paid study was pre-registered, and in each step the agent is checked against a record it cannot edit: tests re-run by the host and, from step 2 on, hidden…</description></item><item><title>Integrating AI Coding Agents Into Scientific Research</title><link>https://arxiv.org/abs/2611.23334</link><guid isPermaLink="false">keryn-headlines-event-ev_7fcc3e382d6a</guid><pubDate>Tue, 17 Nov 2026 01:03:38 GMT</pubDate><category>Proposal</category><category>Agentic SDLC</category><description>New software tools have emerged that enable AI agents to quickly develop code that analyzes data and generates figures, but their mistakes can also harm a researcher&#39;s computer, require extensive resources for validation, and cause the results to be received more skeptically.…</description></item><item><title>Show HN: CoderBaba – An open-source AI Coding Agent with security scanning</title><link>https://coderbaba.si/</link><guid isPermaLink="false">keryn-headlines-event-ev_85b29f96ffee</guid><pubDate>Mon, 16 Nov 2026 15:51:06 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>Hacker News: 5 points, 0 comments.</description></item><item><title>Cursor Supply Chain Attack Started at Fireworks AI</title><link>https://www.securityweek.com/cursor-supply-chain-attack-started-at-fireworks</link><guid isPermaLink="false">keryn-headlines-event-ev_1bf2106fe2f7</guid><pubDate>Mon, 16 Nov 2026 15:45:43 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>The attacker stole an access key from Fireworks and used it to exfiltrate credentials that enabled them to push malicious Claude Code extension copies.</description></item><item><title>Week in review: Secret stealer hits GitHub repos, Claude Code turned into remote-controlled implant</title><link>https://helpnetsecurity.com/2026/11/15/week-in-review-secret-stealer-hits-github-repos-claude-code-turned-into-remote-controlled-implant</link><guid isPermaLink="false">keryn-headlines-event-ev_40fb7e579dae</guid><pubDate>Sun, 15 Nov 2026 09:00:37 GMT</pubDate><category>Roundup</category><category>Agentic SDLC</category><description>Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: What a security leader should tell the CEO about leaked credentials In this Help Net Security interview, Eric Fourrier, CEO of GitGuardian, talks about leaked credentials and…</description></item><item><title>Compositional Exploits in Autonomous Coding Agents: Attack Mechanics and Ephemeral Defense</title><link>https://arxiv.org/abs/2611.19764</link><guid isPermaLink="false">keryn-headlines-event-ev_0a7e7745a783</guid><pubDate>Sun, 15 Nov 2026 03:46:10 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Autonomous coding agents (such as Antigravity, Claude Code, Cursor, and OpenHands) rely on third-party tools to inspect files, run commands, and call external APIs. Today&#39;s security auditors evaluate these tools almost entirely in isolation: static linters and AST checkers…</description></item><item><title>Agent Authenticator: Passkey-Grade, Hardware-Bound Authentication for AI Agents</title><link>https://arxiv.org/abs/2611.19536</link><guid isPermaLink="false">keryn-headlines-event-ev_e85ffd2de96f</guid><pubDate>Sat, 14 Nov 2026 18:00:31 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Trust</category><description>AI coding agents now push code, publish packages, call model APIs and operate cloud resources on behalf of developers. They authenticate to these services with the same bearer secrets humans use, personal access tokens, API keys and package-registry tokens, stored where the…</description></item><item><title>Cursor confirms security breach led to malicious Claude Code extension</title><link>https://bleepingcomputer.com/news/security/cursor-confirms-security-breach-led-to-malicious-claude-code-extension</link><guid isPermaLink="false">keryn-headlines-event-ev_414f52bd2a99</guid><pubDate>Sat, 14 Nov 2026 17:27:25 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>Cursor confirmed it suffered a security breach that let hackers distribute a malicious Claude Code extension through its marketplace after stealing credentials from AI infrastructure provider Fireworks AI. [...]</description></item><item><title>Level Up! AI-Driven Challenges and Adaptations to Code Review Practices</title><link>https://arxiv.org/abs/2611.18939</link><guid isPermaLink="false">keryn-headlines-event-ev_970aa2c89618</guid><pubDate>Fri, 13 Nov 2026 20:47:56 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>AI can write code faster than humans can review, causing a bottleneck and necessitating changes to code review practices. This bottleneck is hotly debated, giving rise to two extreme and orthogonal imagined futures: one where humans continue to be firmly entrenched in the…</description></item><item><title>PrivacyPR: A Dataset of Real-World Privacy Fixes for Benchmarking Privacy-Aware Code Review</title><link>https://arxiv.org/abs/2611.18293</link><guid isPermaLink="false">keryn-headlines-event-ev_cba521808376</guid><pubDate>Fri, 13 Nov 2026 15:18:33 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Trust</category><description>Coding agents increasingly review code, yet whether they catch the privacy defects that laws such as the GDPR require software to avoid remains largely untested, partly because few datasets connect real privacy defects to their accepted fixes and review context. We present…</description></item><item><title>Trained to betray: How post-training supply chain attacks hijack coding agents</title><link>https://microsoft.com/en-us/security/blog/2026/11/12/trained-to-betray-how-post-training-supply-chain-attacks-hijack-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_a871759c1f14</guid><pubDate>Thu, 12 Nov 2026 17:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><category>Models</category><category>SCA</category><description>Microsoft Defender Research and the AI Red Team discuss how poisoned datasets and poisoned model checkpoints can trigger remote code execution and other malicious activities.</description></item><item><title>Best ADLC security tools for teams shipping with coding agents</title><link>https://endorlabs.com/learn/best-adlc-security-tools</link><guid isPermaLink="false">keryn-headlines-event-ev_c083cd78692f</guid><pubDate>Thu, 12 Nov 2026 13:09:58 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><category>Trust</category><description>Which tools secure the agentic development lifecycle? A practical map of agent governance, dependency security, code review, secrets, and runtime controls.</description></item><item><title>AI code review vs SAST: what each catches and how to use both</title><link>https://endorlabs.com/learn/ai-code-review-vs-sast</link><guid isPermaLink="false">keryn-headlines-event-ev_5f5f2477acdc</guid><pubDate>Wed, 11 Nov 2026 15:43:28 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><category>SAST</category><description>AI code review vs SAST: how detection, bug coverage, determinism, false positives, and cost differ, and why combining them catches more real risk.</description></item><item><title>What is AI code review? How it works, what it checks, and where agentic review fits</title><link>https://endorlabs.com/learn/what-is-ai-code-review</link><guid isPermaLink="false">keryn-headlines-event-ev_c7037977af1b</guid><pubDate>Wed, 11 Nov 2026 15:43:28 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><category>SAST</category><description>AI code review uses AI to review pull requests for bugs, quality, and security. Learn how it works, how it differs from SAST and linters, and what agentic review adds.</description></item><item><title>Best code security review tools in 2027 for first-party and AI-generated code</title><link>https://endorlabs.com/learn/best-code-security-review-tools</link><guid isPermaLink="false">keryn-headlines-event-ev_db68513b0f0b</guid><pubDate>Wed, 11 Nov 2026 15:43:28 GMT</pubDate><category>Roundup</category><category>Agentic SDLC</category><description>Compare the best code security review tools of 2026 for pull requests and CI. See how Endor Labs and 8 others review first-party and AI-generated code before merge.</description></item><item><title>ProdPilot: A Layered Model Context Protocol System for Automated Production Readiness and Deployment of Vibe-Coded Projects</title><link>https://arxiv.org/abs/2611.17381</link><guid isPermaLink="false">keryn-headlines-event-ev_726a97c91ac1</guid><pubDate>Wed, 11 Nov 2026 11:36:45 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>AI-assisted development environments turn natural-language requests into working applications, yet the projects they produce often fail on their first deployment for operational reasons: credentials in source files, missing security middleware, no container definition, no…</description></item><item><title>GHA-SmellFix: Taking Action with Agentic LLMs for Detecting and Refactoring GitHub Actions Workflow Smells</title><link>https://arxiv.org/abs/2611.14963</link><guid isPermaLink="false">keryn-headlines-event-ev_ab21a77670c8</guid><pubDate>Wed, 11 Nov 2026 03:16:33 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Continuous integration and continuous delivery workflows are key artifacts in software projects. In GitHub Actions, they encode build, test, release, deployment, and security automation in YAML. These workflows contain configuration smells affecting security, reliability,…</description></item><item><title>Grok Build coding agent leaks oodles of confidential info, some belonging to SpaceXSI</title><link>https://arstechnica.com/security/2026/11/grok-build-coding-agent-leaks-oodles-of-confidential-info-some-belonging-to-spacexsi</link><guid isPermaLink="false">keryn-headlines-event-ev_73b1970125db</guid><pubDate>Tue, 10 Nov 2026 21:12:59 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Researcher finds 35,000 Grok Build apps wide open to anyone who looked for them.</description></item><item><title>AI Coding Agents Leak Thousands of Secrets to GitHub</title><link>https://darkreading.com/cyber-risk/ai-coding-agents-leak-secrets-github</link><guid isPermaLink="false">keryn-headlines-event-ev_8f6d0e7119d4</guid><pubDate>Tue, 10 Nov 2026 18:46:22 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>A command in Claude Code inadvertently saves files with credentials and other sensitive data to public Git repositories, creating security gaps for organizations.</description></item><item><title>Claude Code is accidentally exposing secrets all over GitHub</title><link>https://arstechnica.com/ai/2026/11/claude-code-settings-files-expose-thousands-of-secrets-on-github</link><guid isPermaLink="false">keryn-headlines-event-ev_5697743f6c96</guid><pubDate>Mon, 09 Nov 2026 17:24:26 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>&quot;Do not ask again&quot; prompt leads to passwords, API keys, and more being shared.</description></item><item><title>When Coding Agents Get Arms: Safety Risks Across the Digital-Physical Boundary</title><link>https://arxiv.org/abs/2611.11438</link><guid isPermaLink="false">keryn-headlines-event-ev_78128e431680</guid><pubDate>Sun, 08 Nov 2026 15:02:47 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Advances in spatial understanding have enabled frontier models to power coding agents that control a real robot arm zero-shot and end-to-end, without task-specific demonstrations or retraining. This brings two new safety challenges. First, the extensive knowledge of frontier…</description></item><item><title>Patch correctness in automated program repair: A five-dimensional evidence framework and cross-benchmark audit from classical repair to coding agents</title><link>https://arxiv.org/abs/2611.10837</link><guid isPermaLink="false">keryn-headlines-event-ev_d6ec5b99fbe8</guid><pubDate>Sat, 07 Nov 2026 16:30:54 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Tests guide automated program repair and determine which candidate patches are reported as successful. Repository-level coding agents extend this feedback loop by generating tests and revising code after observing execution results. This technical report introduces a…</description></item><item><title>Sandboxing Untrusted AI Agents at Scale: Google Donates Agent Substrate and gVisor to the CNCF</title><link>https://opensource.googleblog.com/2026/11/sandboxing-untrusted-ai-agents-at-scale.html</link><guid isPermaLink="false">keryn-headlines-event-ev_4a6ab4e42586</guid><pubDate>Thu, 05 Nov 2026 17:00:00 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>by Mofi Rahman , Developer Relations Engineer, and Brandon Royal , Product Manager, Google Cloud When Google introduced Kubernetes twelve years ago, the Pod became the default unit of compute for the cloud. The design assumed a straightforward lifecycle: a program boots, binds…</description></item><item><title>Modernising threat hunting: Using Semantic Search to Find Risky AI Agent Behavior</title><link>https://elastic.co/security-labs/blog/ai-agent-session-monitoring-embedding-models</link><guid isPermaLink="false">keryn-headlines-event-ev_4a059d0fa131</guid><pubDate>Wed, 04 Nov 2026 12:45:02 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>I tested vector search on 7,454 turns of my own agent sessions. One in seven sessions needed review, the agent’s reasoning carried more signal than the prompt, and the stored vectors answer questions nobody thought to ask at ingest.</description></item><item><title>AI supply chain attacks: Old payloads, new entry points</title><link>https://mend.io/blog/ai-supply-chain-attacks-new-entry-points</link><guid isPermaLink="false">keryn-headlines-event-ev_35de003ee95a</guid><pubDate>Tue, 03 Nov 2026 16:13:31 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>SCA</category><description>AI didn&#39;t invent new supply-chain attacks. It opened new paths to old ones.</description></item><item><title>Backporting security fixes at scale with Codex</title><link>https://blog.trailofbits.com/2026/11/02/backporting-security-fixes-at-scale-with-codex</link><guid isPermaLink="false">keryn-headlines-event-ev_6f5ff69cc1db</guid><pubDate>Mon, 02 Nov 2026 12:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>Remediation</category><description>Stable Linux releases lack existing security fixes because backporting them is time-consuming and often thankless work. We’re using Codex to reduce the backporting process time from days to hours per fix, empowering human patch reviewers to rapidly submit fixes for critical…</description></item><item><title>Security notice — malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace</title><link>https://cursor.com/blog/claude-code-extension-security-notice</link><guid isPermaLink="false">keryn-headlines-event-ev_35927976be4f</guid><pubDate>Thu, 29 Oct 2026 12:00:00 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>Today we identified a security incident involving a specific extension in the Cursor VS Code Extension Marketplace: Anthropic.claude-code version 2.1.312 and 2.1.313 for Windows and MacOS. Updated: affected versions.</description></item><item><title>A Supply Chain Attack Put Malicious Anthropic Claude Extensions on Thousands of Devices</title><link>https://wired.com/story/trojan-malware-targets-thousands-of-anthropic-claude-developer-systems</link><guid isPermaLink="false">keryn-headlines-event-ev_3a853fdf7b0c</guid><pubDate>Fri, 30 Oct 2026 20:38:46 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>SCA</category><description>Thanks to a lucky break, flawed code in a tainted Claude extension prevented the attacker from fully worming their way into some 5,000 developer machines.</description></item><item><title>Mend SCA now supports Bun, with reachability built in</title><link>https://mend.io/blog/mend-sca-bun-support</link><guid isPermaLink="false">keryn-headlines-event-ev_46a2789c9f51</guid><pubDate>Thu, 29 Oct 2026 19:10:25 GMT</pubDate><category>Release</category><category>Agentic SDLC</category><category>SCA</category><description>Mend SCA scans bun.lock and bun.lockb with reachability analysis built in.</description></item><item><title>Kerkoporta: Injecting Structured-Output Backdoors in Reasoning Coding Agents</title><link>https://arxiv.org/abs/2611.00138</link><guid isPermaLink="false">keryn-headlines-event-ev_2b95c0816499</guid><pubDate>Thu, 29 Oct 2026 18:41:43 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><category>Models</category><description>Existing large language model (LLM) backdoor attacks operate primarily in chat settings, where the risk is bounded by free-form text outputs. AI agents remove that bound by performing actions and executing commands on the user&#39;s machine. We introduce structured-output backdoor…</description></item><item><title>ExploreRepair: Evidence-Specialized Agents for Vulnerability Repair under Information Constraints</title><link>https://arxiv.org/abs/2610.39899</link><guid isPermaLink="false">keryn-headlines-event-ev_428aa4707ea2</guid><pubDate>Thu, 29 Oct 2026 12:09:12 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><description>Automated vulnerability repair must often operate with incomplete vulnerability information: developers may lack an accurate vulnerability classification or a provided proof-of-vulnerability (PoV) test, yet still need to produce a repair. Without detailed information, a key…</description></item><item><title>StepSecurity Now Inventories AI Agent Plugins in Your GitHub Repositories and on Developer Machines</title><link>https://stepsecurity.io/blog/stepsecurity-now-inventories-ai-agent-plugins-in-your-github-repositories-and-on-developer-machines</link><guid isPermaLink="false">keryn-headlines-event-ev_a26a90ef4338</guid><pubDate>Tue, 27 Oct 2026 14:55:03 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>AI Security</category><description>Inventory the Claude Code and Codex plugins on developer machines and in your GitHub repositories: where each came from, what it bundles, whether it is enabled, and which repositories distribute it.</description></item><item><title>Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models</title><link>https://arxiv.org/abs/2610.35829</link><guid isPermaLink="false">keryn-headlines-event-ev_fa3fde47edca</guid><pubDate>Mon, 26 Oct 2026 17:50:09 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>A vulnerability costs more to fix the later it is discovered. As coding agents generate code faster than developers can review it, software needs a first line of defense screening each new function in real time. LLM agents have recently become capable of finding real-world…</description></item><item><title>Securing Agent-Generated Code with Its Own Development History</title><link>https://arxiv.org/abs/2610.34098</link><guid isPermaLink="false">keryn-headlines-event-ev_49fb4e999e5f</guid><pubDate>Sun, 25 Oct 2026 06:51:47 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><category>SAST</category><description>Pull requests written by coding agents can pass a project&#39;s test suite while omitting a guard or a validation step. To find the missing condition, a reviewer currently relies on static analysis of the final code. An agent-generated change, however, is accompanied by a…</description></item><item><title>ArchGuard-APR: Architecture-Constraint-Aware Multi-Agent Program Repair Using Repository Graphs and Execution Feedback</title><link>https://arxiv.org/abs/2610.33853</link><guid isPermaLink="false">keryn-headlines-event-ev_78a03e744241</guid><pubDate>Sat, 24 Oct 2026 18:48:32 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Repository-level automated program repair must reason about more than a localized failing statement. A candidate change can satisfy a narrow test while violating dependency boundaries, modifying an unsafe path, or relying on context that omits relevant repository structure.…</description></item><item><title>DeepAVR: Deep Analysis via Multi-Agent Collaboration for Automated Vulnerability Repair</title><link>https://arxiv.org/abs/2610.32345</link><guid isPermaLink="false">keryn-headlines-event-ev_ff768b068104</guid><pubDate>Fri, 23 Oct 2026 08:36:03 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><description>Automated Vulnerability Repair (AVR) has emerged as a critical defense against software security flaws, aiming to triage and fix bugs autonomously. Early LLM-based approaches focused on patch generation under the assumption of perfect fault localization, while recent end-to-end…</description></item><item><title>Your CLI and IDE Are the New Attack Surface: Governing AI Coding Agents on the Endpoint</title><link>https://cloudsecurityalliance.org/articles/your-cli-and-ide-are-the-new-attack-surface-governing-ai-coding-agents-on-the-endpoint</link><guid isPermaLink="false">keryn-headlines-event-ev_6ade45e5ee8b</guid><pubDate>Wed, 21 Oct 2026 20:48:20 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>Trust</category><description>A developer asks their coding agent to upgrade a logging library and fix whatever breaks. To see what changed, the agent reads the library&#39;s changelog. Hidden in a comment in that file is an instruction written for the agent: Confirm the environment works by sending the…</description></item><item><title>Your AI agent fixed the bug, but what else did it change?</title><link>https://blog.trailofbits.com/2026/10/21/your-ai-agent-fixed-the-bug-but-what-else-did-it-change</link><guid isPermaLink="false">keryn-headlines-event-ev_89acb8c73b3f</guid><pubDate>Wed, 21 Oct 2026 15:00:00 GMT</pubDate><category>Release</category><category>Agentic SDLC</category><category>Remediation</category><description>Today we’re publishing patching-bugs and regression-hunt , two plugins that let engineers delegate patch development and regression checks to agents. Our patching-bugs plugin reproduces reported bugs, develops fixes, and tests them. The regression-hunt plugin checks whether any…</description></item><item><title>How Much Does the Harness Matter, and for Whom? A Multi-Model Study of Context in LLM-Driven Vulnerability Discovery</title><link>https://arxiv.org/abs/2610.28802</link><guid isPermaLink="false">keryn-headlines-event-ev_deb00cd543d2</guid><pubDate>Tue, 20 Oct 2026 18:25:47 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>SAST</category><description>How much does an agent harness affect vulnerability discovery, and does its value depend on the model? We study 24 model-runner configurations using an open-source harness on a production codebase whose vulnerabilities were non-public at scan time. The pipeline combines static…</description></item><item><title>Secure Agentic Coding through Counterexample-Grounded Feedback</title><link>https://arxiv.org/abs/2610.25916</link><guid isPermaLink="false">keryn-headlines-event-ev_4593e3a727c6</guid><pubDate>Mon, 19 Oct 2026 07:18:34 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>As coding models and agent harnesses become more capable, reliable security verification is becoming a central bottleneck. Improving secure code generation requires a signal whose reported progress remains aligned with actual program behavior under optimization; otherwise, an…</description></item><item><title>How Effectively Do Security Skills Guide Coding Agents in Vulnerability Detection in Linux Kernel?</title><link>https://arxiv.org/abs/2610.25705</link><guid isPermaLink="false">keryn-headlines-event-ev_4800eb9d693c</guid><pubDate>Mon, 19 Oct 2026 02:50:45 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>SAST</category><description>Large language model (LLM) coding agents, such as Claude Code, are increasingly used for security auditing, with task-specific skills providing reusable vulnerability-analysis procedures. However, their effectiveness in detecting real vulnerabilities and the reasons for their…</description></item><item><title>Who Verifies the Builder? Toward Independent Assurance for Agent-Produced Software</title><link>https://arxiv.org/abs/2610.28667</link><guid isPermaLink="false">keryn-headlines-event-ev_a62682f43388</guid><pubDate>Fri, 16 Oct 2026 18:00:10 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Coding agents can now implement a specification, write and run tests, repair failures, and declare the result ready for release. When the same agentic system produces both the software and the evidence of its correctness, a passing test suite is no longer independent evidence:…</description></item><item><title>Your AI agent is sandboxed. Its data still needs guardrails.</title><link>https://mend.io/blog/docker-sandboxes-ai-guardrails</link><guid isPermaLink="false">keryn-headlines-event-ev_c25d962207bb</guid><pubDate>Thu, 15 Oct 2026 16:40:44 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>Docker Sandboxes contains your agent. Mend AI Guardrails checks what it sends.</description></item><item><title>Thinking in Graphs: Neo4j-Backed Exploit-Chain Reasoning for LLM Penetration-Testing Agents</title><link>https://arxiv.org/abs/2610.20902</link><guid isPermaLink="false">keryn-headlines-event-ev_da5cd88de687</guid><pubDate>Wed, 14 Oct 2026 20:33:18 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>LLM coding agents such as Claude Code and OpenAI Codex can drive penetration-testing tools, but they reason about an engagement as a flat list of findings in a context window. Attackers think in graphs: isolated low-severity exposures turn catastrophic once chained through…</description></item><item><title>P2Patch: Exploit-Guided Vulnerability Repair</title><link>https://arxiv.org/abs/2610.20245</link><guid isPermaLink="false">keryn-headlines-event-ev_29fd9e308e8a</guid><pubDate>Wed, 14 Oct 2026 14:19:21 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><description>Large language models are increasingly deployed as autonomous agents for automated program repair, patching software vulnerabilities through multi-step reasoning and tool use. Yet these systems typically validate a patch against the same evidence used to construct it -- a…</description></item><item><title>HPCAgent-Bench: A Framework for Sound Benchmarking of Code-Optimizing Agents, Harnesses, and Skills</title><link>https://arxiv.org/abs/2610.20673</link><guid isPermaLink="false">keryn-headlines-event-ev_de78f66aa74c</guid><pubDate>Wed, 14 Oct 2026 13:22:48 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Coding agents are increasingly used to optimize code, but existing benchmarks fix one accelerator, programming model or harness, so they cannot isolate what each component contributes or show whether added skills and tools repay their cost. We present HPCAgent-Bench with 689…</description></item><item><title>Visual Studio Copilot and SSMS Remote Code Execution (CVE-2026-103572)</title><link>https://embracethered.com/blog/posts/2026/visual-studio-copilot-remote-code-execution</link><guid isPermaLink="false">keryn-headlines-event-ev_a279a97ad1e6</guid><pubDate>Tue, 13 Oct 2026 22:00:40 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>AI Security</category><category>Remediation</category><description>A . is all you need! Today Microsoft shipped the patch for a vulnerability in Visual Studio Copilot that I reported in August. Microsoft classified it as Important severity Remote Code Execution vulnerability. At the root cause it allows Copilot in Visual Studio or SQL Server…</description></item><item><title>AI coding agents expose developer screenshots on GitHub repos</title><link>https://reversinglabs.com/blog/ai-coding-agents-expose-developer-screenshots-on-github</link><guid isPermaLink="false">keryn-headlines-event-ev_24ccdde62b1c</guid><pubDate>Tue, 13 Oct 2026 15:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>PixelLeak exposures included many sensitive documents, highlighting why inventories of what agents can reach and publish are needed.</description></item><item><title>Auditing the auditor: Unmasking Apex Accelerator</title><link>https://elastic.co/security-labs/threat-command/apex-accelerator-malicious-npm-packages</link><guid isPermaLink="false">keryn-headlines-event-ev_a2254c57a64b</guid><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>REF8801 exposes how Apex Accelerator combines convincing Web3 services with malicious npm packages, attacker-controlled AI infrastructure, and cross-platform malware delivery.</description></item><item><title>EviRepair: Evidence-Driven Agents for Reliable Vulnerability Repair</title><link>https://arxiv.org/abs/2610.13491</link><guid isPermaLink="false">keryn-headlines-event-ev_d73ff45a31eb</guid><pubDate>Fri, 09 Oct 2026 12:22:43 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><description>Recent advances in large language models (LLMs) have enabled agent-based approaches for repository-level automated vulnerability repair (AVR), showing promising repair capabilities. However, an important issue has received limited attention: how can we ensure that the evidence…</description></item><item><title>Frontier models found the vulnerabilities. Only the attacker found the chains.</title><link>https://snyk.io/blog/frontier-models-vulnerabilities-attacker-chains</link><guid isPermaLink="false">keryn-headlines-event-ev_b87055221917</guid><pubDate>Wed, 07 Oct 2026 04:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>SAST</category><description>Static analysis found the flaws, but only live attack testing proved how they could be chained into breaches. A comparison of Evo COS, Claude Security, and Claude Code Security.</description></item><item><title>ReviewBench: An open benchmark for AI code review</title><link>https://github.blog/ai-and-ml/github-copilot/reviewbench-an-open-benchmark-for-ai-code-review</link><guid isPermaLink="false">keryn-headlines-event-ev_ef54545eba84</guid><pubDate>Mon, 05 Oct 2026 15:59:40 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><description>We’re launching ReviewBench, a benchmark for code review agents built on representative GitHub pull requests, multi-source ground truth, calibrated evaluation, and production-aligned metrics.</description></item><item><title>Where Did the Repair First Go Wrong? Localizing the Origins of Silent Failures in Agentic Vulnerability Repair</title><link>https://arxiv.org/abs/2610.06163</link><guid isPermaLink="false">keryn-headlines-event-ev_a5410677a281</guid><pubDate>Mon, 05 Oct 2026 11:31:16 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>Remediation</category><description>Localizing where an LLM-based agent first fails to uphold security during a repair can show which stage of its workflow needs an additional safeguard. This is difficult for silent failures, which are patches that pass syntactic and functional checks but still contain a security…</description></item><item><title>Why AI Coding Agents Keep Writing Broken Access Control</title><link>https://snyk.io/blog/ai-coding-agents-broken-access-control</link><guid isPermaLink="false">keryn-headlines-event-ev_00c17969330f</guid><pubDate>Thu, 01 Oct 2026 04:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><description>AI coding agents can produce authorization logic that compiles and passes review while exposing one tenant’s data to another. Learn why broken access control is difficult to detect and how to prevent it.</description></item><item><title>From SELECT to SYSADMIN with SQL Copilot (CVE-2026-65669)</title><link>https://embracethered.com/blog/posts/2026/from-select-to-sysadmin-sql-copilot-bluehat-asia</link><guid isPermaLink="false">keryn-headlines-event-ev_57e4573a0ecb</guid><pubDate>Wed, 30 Sep 2026 21:00:40 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><description>Two weeks back I presented at BlueHat Asia 2026 about my research on Microsoft’s Copilot in SSMS, the SQL Server Management Studio. This post is a write up about the talk, which covered CVE-2026-65669 , a SQL Server Elevation of Privilege Vulnerability rated critical by…</description></item><item><title>SecProbe: Adaptive Evaluation of Coding Agents on Cybersecurity Vulnerabilities</title><link>https://arxiv.org/abs/2609.33763</link><guid isPermaLink="false">keryn-headlines-event-ev_99a98665c7ac</guid><pubDate>Sun, 27 Sep 2026 17:01:35 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Assessing cybersecurity vulnerability awareness in coding agents requires evaluations that reveal capability gaps and remain informative as models evolve. Static benchmarks offer fixed coverage and difficulty, while scarce vulnerable repositories and costly expert authoring…</description></item><item><title>Auditing in the age of (good enough) AI</title><link>https://blog.trailofbits.com/2026/09/18/auditing-in-the-age-of-good-enough-ai</link><guid isPermaLink="false">keryn-headlines-event-ev_da19e5188bda</guid><pubDate>Fri, 18 Sep 2026 11:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>SAST</category><description>Security firms have published numerous blog posts describing how they pointed their agent harness at a codebase and found dozens of bugs ( we’re one of them ). However, these posts tend to focus on agentic code review, which is just one aspect of how we use AI in our security…</description></item><item><title>Agentic AI permissions: A core problem — but not the only one</title><link>https://reversinglabs.com/blog/agentic-ai-permissions</link><guid isPermaLink="false">keryn-headlines-event-ev_912d4cf4c5af</guid><pubDate>Thu, 17 Sep 2026 15:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><description>Controlling coding agent overpermissioning is key to security. But recent frontier AI incidents show that problems don’t stop there.</description></item><item><title>Adversarial Testing of Automated Program Repair Agents for Security Vulnerabilities</title><link>https://arxiv.org/abs/2609.15963</link><guid isPermaLink="false">keryn-headlines-event-ev_e552b0d352e0</guid><pubDate>Mon, 14 Sep 2026 17:53:49 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><description>Software agents with Large Language Models (LLMs) are designed for Automated Program Repair (APR) tasks, raising the possibility that, in the near future, APR agents will fix bugs automatically without much human intervention. Can we trust an APR agent to produce both…</description></item><item><title>Just a rumour of a bug is enough to find a security exploit these days</title><link>https://simonwillison.net/2026/Aug/28/just-a-rumour-of-a-bug</link><guid isPermaLink="false">keryn-headlines-event-ev_61cbc9ef4277</guid><pubDate>Fri, 28 Aug 2026 22:12:02 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><category>Remediation</category><description>Just a rumour of a bug is enough to find a security exploit these days Anil Madhavapeddy is a professor of computer science at Cambridge and a core maintainer of the OCaml compiler. In this somewhat alarming post he reports that security issues in OCaml projects are seeing…</description></item><item><title>Breaking Claude Code Opus 5 Auto Mode</title><link>https://embracethered.com/blog/posts/2026/breaking-claude-code-opus-5-and-automode</link><guid isPermaLink="false">keryn-headlines-event-ev_b34e3c0648da</guid><pubDate>Thu, 27 Aug 2026 04:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>In this post, we explore how a simple website summary request hijacks Claude Code Opus 5 in Auto Mode and achieves code execution with 60-80% attack success rate using a small sample size. This is interesting because a third-party evaluation commissioned by Anthropic showed a…</description></item><item><title>13 million tool calls: auditing every AI coding agent action with Elastic Agent</title><link>https://elastic.co/security-labs/blog/ai-coding-agent-audit-cursor-hooks</link><guid isPermaLink="false">keryn-headlines-event-ev_201a4303fd45</guid><pubDate>Tue, 11 Aug 2026 00:00:00 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><description>Cursor hooks and Elastic Agent capture every tool call, shell command, file read and MCP request as structured events you can hunt with ES|QL.</description></item><item><title>Auto mode is now the default in Claude Code for Pro, Max, and Team plans</title><link>https://simonwillison.net/2026/Aug/8/auto-mode</link><guid isPermaLink="false">keryn-headlines-event-ev_4903d8624c38</guid><pubDate>Sat, 08 Aug 2026 22:36:03 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><category>AI Security</category><description>Auto mode is now the default in Claude Code for Pro, Max, and Team plans Anthropic are really confident in Claude Code&#39;s auto mode , to the point that they are making it the default setting for new sessions in most Claude Code plans starting on August 14th. This was one of the…</description></item><item><title>Living off the coding agent: Two tales of tunnels and LaunchAgents</title><link>https://elastic.co/security-labs/threat-command/coding-agent-launchagent-tunnel-detection</link><guid isPermaLink="false">keryn-headlines-event-ev_876972dd96da</guid><pubDate>Fri, 07 Aug 2026 00:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><description>Agent-parented reverse tunnels and LaunchAgents can expose a local admin app to the internet. Endpoint still needs to treat that as high severity even when the activity looks like vibe-coded ops, not confirmed malware.</description></item><item><title>A First Look at Evo Agentic AppSec: Agentic Remediation and Malicious Code Defense</title><link>https://snyk.io/blog/remediation-agent-malicious-code-defense</link><guid isPermaLink="false">keryn-headlines-event-ev_088987fd1131</guid><pubDate>Tue, 04 Aug 2026 04:00:00 GMT</pubDate><category>Announcement</category><category>Agentic SDLC</category><category>Remediation</category><category>SCA</category><description>Explore Snyk’s first Agentic AppSec capabilities: an autonomous Remediation Agent that fixes vulnerabilities and Malicious Code Defense that blocks risky packages before they ship.</description></item><item><title>Escaping Linux Sandboxes via PipeWire (CVE-2026-5674)</title><link>https://embracethered.com/blog/posts/2026/pipewire-flatpak-linux-sandbox-escape-cve-2026-5674</link><guid isPermaLink="false">keryn-headlines-event-ev_e9e2b2ad3349</guid><pubDate>Thu, 30 Jul 2026 16:00:00 GMT</pubDate><category>Analysis</category><category>Agentic SDLC</category><description>This post walks through a sandbox escape from a Flatpak application via PipeWire. The vulnerability was discovered using my automated research pipeline with Claude Code and Opus 4.6 back in April 2026. It was an exciting find, as this was the first bug I submitted to Red Hat.…</description></item><item><title>How we use /goal to find bugs in Patch the Planet</title><link>https://blog.trailofbits.com/2026/07/28/how-we-use-goal-to-find-bugs-in-patch-the-planet</link><guid isPermaLink="false">keryn-headlines-event-ev_a9322bd2c984</guid><pubDate>Tue, 28 Jul 2026 11:00:00 GMT</pubDate><category>Guide</category><category>Agentic SDLC</category><category>Remediation</category><category>SAST</category><description>Codex’s /goal feature amplifies bug hunting, but getting good results requires the right prompt, the right scope, and the right number of outcomes per run. For Patch the Planet , our joint initiative with OpenAI to find and fix bugs in open-source software, we pointed Codex at…</description></item></channel></rss>