This Week's Highlights — issue 019, November 6, 2026
THE CYBER TOOLCHAIN
NO. 019
The best tooling updates that shipped this week.
The week in numbers · Oct 30–Nov 5
Changes per week, last 12 weeks · Oct 30–Nov 5
Changes we read in that week's dailies, including docs and API changes. A gap means no daily was published.
Every feature release from the 1311 tools on our watchlist goes into the daily newsletter. Once a week we read the whole field side by side and grade each release from 1 to 5 on three things, weighted equally:
- Depth. How complete is the best new capability?
- Impact. How much does it change what you can do?
- Novelty. Has any other tool done this before? We check every capability against the history of every tool we track. Features most tools already have earn nothing here. Being first earns the most.
A tool is judged on everything it shipped that week, so a project that releases daily gets credit for the sum — and still only takes one slot. We would rather run a short list than a padded one.
Focus: What features shipped and what problems they solve, plus launches, packaging and where the market is moving.
Leaves out: Command-line flags, API-only deltas, usage examples, and releases that are only refinements.
Focus: New security controls, new APIs, and product improvements you can put to work, plus anything an upgrade breaks.
Leaves out: Launch and marketing news, governance, pricing, and documentation changes with no operational effect.
Focus: Dashboards, reporting, and anything you can quantify: metrics, coverage, risk scores, compliance impact, vendor moves.
Leaves out: Settings, flags, API and command-line detail, library bumps, and minor releases with no reporting or posture angle.
Focus: API, SDK and MCP changes, new libraries, breaking changes in what you import, and cyber tooling.
Leaves out: Governance, SOC operations, and vendor launches with no developer surface.
This issue came out before editions were ranked, so it shows the Standard edition in issue order.
The picks in full
SentinelOne's Singularity platform protects endpoints, cloud, identity and data with AI-driven detection, Purple AI and an agentic SOC layer.
Singularity Agentic Detection and Response builds a live inventory of AI agents, tools, connectors and MCP servers with their permissions, and can contain an individual agent without halting the wider workflow — a control for an attack surface most EDRs do not yet touch. It also links AI-driven actions to subsequent processes to flag when an agent deviates from its task.
SentinelOne launched Singularity Agentic Detection and Response (ADR) to discover, monitor, and contain AI agents across the enterprise, and added Anthropic's Claude Mythos 5.1 model to its Wayfinder Frontier AI Services for exploitability-validated vulnerability assessment.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
what's the number?
Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.
- depth0–40 what it does and how it works or what changed
- specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
- actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point
Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.
ADR builds an inventory of AI agents, tools, connectors, and Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. (MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) servers along with their permissions and associated users, supporting agents built on Salesforce Agentforce, Microsoft Copilot Studio, OpenAI ChatGPT workspace agents, and Anthropic Claude managed agents.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ It monitors agent activity through event traces linking AI-driven actions to subsequent system processes to detect deviation from assigned tasks, lets security teams set controls on which agents can operate and which resources they can access, and can contain an individual AI agent without stopping the wider workflow. ADR detects and blocks threats including prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools. and credential exfiltration before agent actions execute, and integrates agent activity into the Singularity Platform's existing investigation and response workflows.
Wayfinder Frontier AI Services now includes Anthropic Claude Mythos 5.1, a restricted-access model, alongside existing OpenAI and Anthropic models to improve vulnerability discovery and exploitability confirmation. Findings are filtered through exploitability testing so customers receive validated results and prioritised remediation guidance rather than raw vulnerability lists.
Orbit indexes projects, pipelines, security findings, and ownership into one relationship graph you can query in plain language to map a CVE's blast radius across a whole group, and external agents such as Claude Code and Codex reach it over MCP. It replaces manual correlation with a live, queryable context graph of the software lifecycle.
GitLab shipped Orbit, a live relationship graph connecting code, merge requests, pipelines, vulnerabilities, work items and ownership across projects, reaching general availability for Premium and Ultimate customers with plain-language blast-radius queries, MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface./REST agent access, exact code search alongside the graph, and usage-based query-pack pricing.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
what's the number?
Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.
- depth0–40 what it does and how it works or what changed
- specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
- actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point
Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.
MCP-compatible agents — Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Antigravity — connect to the Orbit context graph by running glab auth login then glab orbit setup; custom agents and internal tooling can connect through REST.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Agents on Duo Agent Platform query Orbit natively with nothing to configure, and GitLab's built-in Security Analyst Agent now uses Orbit directly to traverse the graph and answer blast-radius questions from indexed project data.
$ glab auth login
glab orbit setup
GitLab adds Orbit, a live, queryable graph connecting code, merge requests, pipelines, vulnerabilities, work items and ownership across all projects in a top-level group — indexing projects, merge requests, pipelines, work items, security findings and source code into one graph. It indexes every push, merge and deploy in real time, handling more than 1.5 million projects and 8 TB of graph data in under 2 hours at GitLab scale, and is now generally available for Premium and Ultimate customers on GitLab.com, with Self-Managed and GitLab Dedicated support coming in the following weeks and a free 30-day trial available through account teams.
Exact code search at the top-level group now supports regular expression mode for finding container base images, e.g. lang:Dockerfile ^FROM.*php or file:composer\.json$, used alongside Orbit graph queries.
$ lang:Dockerfile ^FROM.*php
Query blast radius in plain language via GitLab Duo Agentic Chat — for example asking it to map the blast radius of PHP in a group, listing every project with PHP code grouped by subgroup with owners, or to collect every container scanning finding for specific CVEs across the group from the vulnerability report (this requires Ultimate and the Security Manager role).
Orbit is sold as query packs — a shared pool per organization — priced at $30 in GitLab Credits per user per month for 1,000 queries, with overages billed at 33 queries per GitLab Credit (about $0.03 per query).
Orbit access follows existing GitLab roles, so people and agents see only what they are permitted to see, and the graph queries behind each answer can be audited.
Engineers can query the Orbit graph through the Data Explorer.
Adaptive Detonation adds an AI agent that runs a sample, identifies what blocked execution — an evasion check, missing prerequisite, or failure — then supplies missing launch arguments, adjusts the guest clock or environment, and re-runs the unchanged sample to surface behavior the first detonation missed. It automates defeating sandbox evasion that previously stalled analysis entirely.
Joe Sandbox introduces Adaptive Detonation, an AI agent built into Joe Sandbox AI that diagnoses why a sample stalled and actively intervenes — supplying missing arguments, adjusting the guest clock, or preparing files and dependencies — before re-running the sample to expose behavior the initial pass missed.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
what's the number?
Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.
- depth0–40 what it does and how it works or what changed
- specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
- actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point
Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.
Joe Sandbox adds Adaptive Detonation, an AI agent that runs a sample, identifies what blocked execution (evasion check, missing prerequisite, or failure), applies a targeted intervention, and re-analyzes the sample under the new conditions.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Interventions include supplying missing launch arguments — e.g. recovering and injecting the required S1=1 MSI installer property — without modifying the sample package; adjusting the guest machine's date or advancing its clock (demonstrated by a two-day advance to bypass a file-age loader delay) before re-running the unchanged sample in Joe Sandbox Cloud; and running preparation code to create required files, directories, or registry values, or install a missing dependency such as Python in a location the sample already searches. Analysis reports now include the agent's findings: what blocked execution, the runtime and code evidence, the recommended action, alternative explanations, and remaining uncertainty. It is available as part of Joe Sandbox AI and has been running in Joe Sandbox Cloud Basic.
Sysdig AI Defense is a runtime layer that blocks unvetted MCP servers, skills, and plugins before they run and holds risky credential reads for review, with every agent session recorded as an immutable, audit-grade action graph tied to an accountable human. Built from runtime telemetry across Claude Code, Codex, Copilot, and Cursor, it lets a team stop and prove agent actions from endpoint to cloud rather than just inventory them.
Sysdig Secure's biggest addition this window is AI Defense, a new runtime security and governance layer for AI agents covering inventory, policy enforcement, audit trails, and cost attribution; alongside that, Vulnerability Findings gained grouped notifications and the CLI Scanner reached v1.32.0 with new OS support and detection improvements.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
what's the number?
Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.
- depth0–40 what it does and how it works or what changed
- specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
- actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point
Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.
When scanning images against on-premises collectors 7.5.0 to 7.9.x that contain standalone binary applications, the CLI Scanner now excludes applications from the SBOMSBOMA software bill of materials, an inventory of the components and versions inside an application. It helps teams find affected software when a dependency has a vulnerability or licensing issue. and scan results; upgrading the on-premises backend to 7.10.0 or later restores application inclusion.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
Sysdig CLI Scanner v1.32.0 adds detection of Claude Code installed via its native installer during image scanning; adds support for scanning images on SUSE Linux Enterprise Server 16, openSUSE Leap 16, and Ubuntu 26.10 (including end-of-life information for SUSE Linux Enterprise Server 16); marks vulnerabilities with a known exploit as exploitable even when the exploit has no publication date or reference link; and improves PHP detection in official Docker Hub php images by also reading the PHP_VERSION environment variable.
AI Defense enforces runtime policy against agent actions, blocking unvetted MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers, skills, and plugins before they run, and holding or blocking risky credential reads and behavior drift for human review, with exception paths for developers.
Vulnerability Findings automations can now group findings by resource or image, sending one Slack or email notification — or creating one Jira ticket — per resource or image instead of per finding (Preview).
AI Defense provides a live inventory of AI agents, harnesses, models, MCP servers, skills, and plugins built from runtime telemetry, with every agent traceable to an accountable human across platforms including Claude Code, Codex, GitHub Copilot, and Cursor.
After upgrading, scans of Fedora images may show additional vulnerabilities because Fedora subpackages are now matched against their source package instead of the previous matching logic.
AI Defense attributes every model call to an agent, team, and individual — including calls that bypass an AI gateway — to surface real-time AI spend and identify where premium models are doing routine work.
AI Defense records every agent session as an immutable, audit-grade action graph showing each step in order, which policies were applied, when, and who signed off.
AI Defense ships with policy templates that include rules mapped to OWASPOWASPA nonprofit foundation that publishes open standards, guides, and ranked vulnerability lists—most famously the OWASP Top 10—giving cyber tools a vendor-neutral benchmark to test or report against. Top 10 for LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. Applications, with policies starting in alert mode and moving to enforcement when teams are ready.
Sysdig AI Defense is generally available for macOS and Windows.
- New compliance framework CIS Kubernetes Benchmark v2.1.0 control 1.2.1 node audit
- New compliance framework CIS Kubernetes Benchmark v2.1.0 control 4.1.11 node audit
- !After upgrading, scans of Fedora images may show additional vulnerabilities because Fedora subpackages are now matched against their source package instead of the previous matching logic.
- !When scanning images against on-premises collectors 7.5.0 to 7.9.x that contain standalone binary applications, the CLI Scanner now excludes applications from the SBOM and scan results. Upgrade the on-premises backend to 7.10.0 or later to restore application inclusion.
AI Assisted Authentication uses an LLM to drive a real browser through multi-step logins that Browser Based Authentication heuristics cannot handle, then converts the successful session into a reusable Zest replay script — closing a long-standing gap where complex auth flows left apps unscannable. The same release adds OAuth2 client_credentials and password-grant authentication.
OWASP ZAP's headline addition is AI Assisted Authentication, which uses an LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. to drive a real browser through complex multi-step login flows and auto-generate a reusable Zest script; the same window also added OAuth2OAuth2An open authorization framework (RFC 6749) that lets a service grant limited access to a user's resources on another service without exposing credentials, giving cyber tools a standard way to authenticate against third-party APIs. authentication, an in-browser Zest script editor, LLM provider configuration, and a passive rule for server-side technology CVEs.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
what's the number?
Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.
- depth0–40 what it does and how it works or what changed
- specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
- actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point
Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.
A new ai authentication method in the Automation Framework authentication block (parameters loginPageUrl, browserId, loginPageWait, optional hint) uses an LLM to drive a real browser through multi-step login flows that Browser Based Authentication heuristics cannot handle, then automatically converts the successful session into a reusable Client Script Authentication Zest script.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ It can also be triggered programmatically via the aiAssistedAuthentication action in the authentication API, steered with free-text hint guidance (e.g. 'Set the domain dropdown to Project1 before clicking Next.'), monitored step-by-step in the new AI Assisted Auth tab inside the LLM Chat tab, and tried interactively via Tools > Authentication Tester (or Ctrl+T) by selecting AI Assisted as the Auth Method.
env:
contexts:
- name: "ai-auth-context"
urls:
- "https://www.example.com"
authentication:
method: "ai"
parameters:
loginPageUrl: "https://www.example.com/login"
browserId: "firefox-headless"
loginPageWait: 5
hint: "Set the domain dropdown to Project1 before clicking Next."
verification:
method: "autodetect"
sessionManagement:
method: "autodetect"
users:
- name: "test-user"
credentials:
username: "[email protected]"
password: "${AUTH_PASSWORD}"
jobs:
- type: llm-config
parameters:
defaultProvider: "my-provider"
defaultModel: "my-model"
providers:
- name: "my-provider"
type: "OPENAI_COMPATIBLE"
apiKey: "${LLM_API_KEY}"
endpoint: "https://llm.example.com/v1"
models:
- "my-model"
- type: "requestor"
parameters:
user: "test-user"
requests:
- url: "https://www.example.com/"
method: "GET"
responseCode: 200
ZAP's Insights add-on (v0.7.0) adds an exitAutoLevel option (accepting medium or high) to the insights Automation Framework job, preferred over the now-secondary exitAutoOnHigh field.
- type: insights
parameters:
exitAutoLevel: high
ZAP's Client Side Integration add-on (v0.32.0) adds a runOnlyIfModern option to the spiderClient Automation Framework job to run the Client Spider only when a Modern Web Application alert has been raised, and now stands in for the AJAX Spider's spiderAjax automation job and ajaxSpider API using the Client Spider when the AJAX Spider add-on is not installed.
- type: spiderClient
parameters:
runOnlyIfModern: true
ZAP's Authentication Helper add-on (v0.43.0) adds an OAuth2 authentication method supporting the client_credentials and password grant types, configurable via the desktop UI, the Automation Framework (oauth2 method, added in Automation Framework v0.61.0), and the API.
ZAP's Authentication Helper add-on (v0.43.0) adds type and count parameters to the diagnostics automation job, retaining only the last failed attempt with auth_on_failure or a rolling window of failures with auth_failure_rolling.
ZAP's LLM Support add-on (v0.1.0) adds an llm-config Automation Framework job to configure an LLM provider, specifying defaultProvider, defaultModel, and a providers list with type, apiKey, endpoint, and models fields; it underpins AI Assisted Authentication.
ZAP's Browser Extension (v0.3.0, included in Client Side Integration v0.34.0) adds a Script Viewer that lets you see, edit, replay (including step-by-step), load, and save a Zest script without leaving the browser.
ZAP's Passive scanner rules (v77) add a Server Side Technology Version with Known Vulnerabilities rule that checks technology versions disclosed in response headers against known vulnerabilities, initially covering PHP versions via php-src security advisories.
ZAP's Network add-on now notifies other add-ons when an outbound proxy is enabled (v0.30.0) and when requests fail to send, such as DNS errors, connection refused, or timeouts (v0.32.0).
ZAP's Automation Framework (v0.61.0) now allows defining how long a plan can run.
Where each pick stands · Oct 30–Nov 5
Tools with changes in each pick's category over the last 12 weeks, including the pick itself.
Each pick's release rhythm · Oct 30–Nov 5
Changes per week for the last 12 weeks. The bright bar is this week; a dash marks a week with no daily.
Which categories moved · Oct 30–Nov 5
The thick bar is this week. The thin bar is the average of the four weeks before, counting only weeks with a daily.
Who shipped the most · Oct 30–Nov 5
Changes per tool this week. Volume is not a quality score.