This Week's Highlights — issue 008, August 21, 2026

Plerion now builds an external access grant inventory for every tenant, analyzing 97 AWS resource types across four sharing mechanisms: resource policies, role trust policies, Resource Access Manager (RAMRAMWorking memory that holds programs and data while a computer is running. Its contents usually disappear when power is removed.) shares, and cross-account permissions set directly as an attribute. Grants can be exported to a spreadsheet for offline review, or reviewed end-to-end in-product: assign an accountable team, record a Keep, Remove, Review later, or Trust until review decision, add notes, and set a future review date, with a full audit history per grant and a Past due preset to surface overdue reviews. A finding is raised for every asset granting access to an external principal not confirmed as trusted, with trusted principals configured under Settings > Profiles > Trusted principals.

Plerion rebuilt 40 compliance frameworks from scratch against the current detection catalog, producing 15,250 control-to-detection mappings across AWS, Azure, GCP, and now Kubernetes (29 frameworks). Eighteen frameworks were promoted to their publisher's latest edition, including PCI DSSPCI DSSThe Payment Card Industry Data Security Standard, security requirements for protecting payment account data. It covers organizations that store, process or transmit that data and relevant service providers. 4.0.1, NIST SP 800-53 Rev 5 Release 5.2.0, CSA CAIQ v4.1, UK NCSC CAF 4.0, and FedRAMPFedRAMPA US government program that standardizes cloud security authorization, letting vendors get a single assessment accepted across federal agencies instead of repeating the process for each one. Rev 5 CR26 baselines. CISCISA nonprofit that publishes consensus-based configuration benchmarks and controls for hardening operating systems, cloud environments, and software; cyber tools use it as a ready-made compliance baseline. Amazon Web Services Foundations Benchmark v7.0.0 was added with all 70 recommendations mapped, including six new AWS Organizations controls (centralized root access, authorization guardrails, organizational unit structure, delegated administrators) and a check for reaching AWS services over VPCVPCA logically isolated virtual network inside a public cloud provider's infrastructure, letting tools deploy resources with private addressing, controlled routing, and firewall rules without managing physical hardware. endpoints, each tagged CIS Level 1 or Level 2. Twenty-three superseded framework editions were retired — including PCI DSS 4.0 and 3.2.1, NIST CSFNIST CSFThe NIST Cybersecurity Framework, a set of outcomes for managing cybersecurity risk. Organizations use it to describe their current posture, priorities and improvement goals. v1.1, ISO/IEC 27001:2013, CIS AWS Foundations Benchmark v4.0.1, and FFIEC CAT (with no successor) — removing them from the framework list while preserving existing mappings; as a result, compliance posture percentages will shift in both directions across all 40 rebuilt frameworks.

New API endpoints let teams read, write, and clear detection exemptions on any detection without using the platform UI. Exemption matching supports asset name, tags, region, principal, or API Gateway route as criteria, and existing findings on the dashboard are re-evaluated immediately when an exemption is written rather than waiting for the next scan. Profiles should be listed via the API first, since the default profile only covers integrations with no profile of their own.

Endor Labs published two complete Endor Patch examples in the endor-patches repository with reproducible builds, attestations, and auditable diffs. Patched JARs in patches/<library>/artifacts/ can be installed to a local Maven repository via mvn install:install-file and declared as a dependency using the -endor-YYYY-MM-DD version string, with transitive override support via <dependencyManagement> in Maven or resolutionStrategy in Gradle; the attestation JSON can be inspected to verify security diffs against upstream commit SHAs, and the artifact can be rebuilt from source and checksum-verified.

$ cat patches/com.fasterxml.woodstox/woodstox-core/security_attestation.json | jq '.[] | select(.fix_type == "FIX_TYPE_VULN")'

$ # Extract the source tarball, then build with Docker and Bazel per the included README.md
tar -xf patches/com.fasterxml.woodstox/woodstox-core/source.tar.gz
cd woodstox-core-patched && docker run --rm -v "$PWD":/build <hermetic-image> bazel build //...
sha256sum output/woodstox-core-6.3.1-endor-2024-09-03.jar
diff <(sha256sum output/woodstox-core-6.3.1-endor-2024-09-03.jar | awk '{print $1}') SHA256SUM.txt

$ mvn install:install-file \
  -Dfile=patches/com.fasterxml.woodstox/woodstox-core/artifacts/woodstox-core-6.3.1-endor-2024-09-03.jar \
  -DgroupId=com.fasterxml.woodstox \
  -DartifactId=woodstox-core \
  -Dversion=6.3.1-endor-2024-09-03 \
  -Dpackaging=jar

Full C language support was added to Endor Labs AI SAST, with buildless scans that run in the IDE and on pull requests with no compiler hook or reproducible build required. It pairs a deterministic call-graph and reachability engine with 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. reasoning to trace data flow across functions and files, catching buffer overflows, use-after-free, integer overflows, and other cross-function memory-safety flaws that pattern-based rules miss. C SAST runs alongside C SCASCASoftware composition analysis, identifying third-party dependencies and checking them for known vulnerabilities and license issues. It covers software a team uses rather than writes itself. in the same platform, covering first-party C code and its open-source dependencies together, and is also surfaced through AURI by Endor Labs, giving an independent security harness outside the coding agent with policy enforced across any agent, model, or CI stage.

Adds .afullhistorycloneoverridepolicy.toml (TOML, schema_version = 2) at the repo root to declaratively configure clone optimization, including [git_config], [fetch], [sparse_checkout], exclude_extensions, conemode, includes, excludes, case_variants, and [policyinfo] fields (description, OptimalForAIAgents). A self-contained Go binary, git-clone-override-policy, intercepts bare git clone <URL> calls and runs a fixed 13-step shallow-and-partial-clone sequence before applying the repo's policy. [fetch] supports flags like --depth=1 and --filter=blob:none; [sparse_checkout] supports conemode = 'auto' (resolving to cone or no-cone at runtime), includes/excludes, and exclude_extensions covering 30+ binary file types; [git_config] scopes large-transfer tuning (core.compression, http.postBuffer, http.lowSpeedLimit, http.lowSpeedTime, pack.windowMemory, pack.packSizeLimit, pack.threads) to the cloned repo only, without touching global git config. The binary runs on Linux, macOS, and Windows for amd64 and arm64 (CMD, PowerShell, Git Bash, WSLWSLWindows Subsystem for Linux, a Microsoft-built compatibility layer that lets a Linux environment run directly on Windows without a virtual machine, giving cyber tools access to Linux binaries, filesystems, and syscalls on a Windows host.) with only a git binary on PATH required, and is bypass-safe: clones already carrying --depth, --filter, or any other flag, and repos without a policy file, pass through unchanged.

⤢

▶ ⤢

▶ ⤢

# .afullhistorycloneoverridepolicy.toml at repo root
schema_version = 2

[policyinfo]
description = "Optimized clone for CI and AI agents"
OptimalForAIAgents = true

[git_config]
"core.compression" = 0
"http.postBuffer" = "1024M"
"pack.threads" = 4

[fetch]
flags = ["--depth=1", "--filter=blob:none"]

[sparse_checkout]
enabled = true
conemode = "auto"
includes = ["/*"]
exclude_extensions = [
  "png", "jpg", "gif", "svg", "pdf",
  "zip", "tar", "gz", "mp4", "exe",
  "dll", "woff", "woff2", "ttf", "psd"
]
case_variants = "both"

$ ./git-clone-override-policy https://example.com/group/my-repo.git

The Flow Creator agent, accessible via GitLab Duo Agentic Chat, converts plain-language descriptions into complete, runnable Flow Registry YAML definitions with no YAML schema knowledge required. It reads current Flow Registry documentation before every response, applies rules encoding known failure patterns, and runs a pre-output checklist catching issues like missing project_id, broken human-in-the-loophuman-in-the-loopHuman in the loop, involving a person in an otherwise automated workflow. The person can review, approve, correct or supply information before work continues. gates, wrong sends_response_to targets, and missing stopping instructions. Generated definitions can be pasted into the configuration editor when creating a flow from a project or the AI Catalog with private, restricted, or public visibility; enabling a flow still requires Maintainer role or higher, Group Owners and instance administrators can disable Custom Flows entirely, and every flow runs under a scoped service account with composite identity bounded to the top-level group.

Adds a bulk-select UI in the vulnerability report to run SAST False Positive Detection and Agentic SAST Vulnerability Resolution against multiple findings at once, replacing one-at-a-time manual triage. False Positive Detection analyzes surrounding code and returns a confidence score and explanation; SARIFSARIFStatic Analysis Results Interchange Format, a standard for sharing code-analysis findings, including locations and rule details. It lets scanners report results to a common review interface.-imported findings from third-party scanners into the unified vulnerability report are also eligible for these bulk agentic flows. Pipelines can run both automatically on Critical and High severity new findings so triage and fix MR creation happen before a developer opens the finding, and bulk jobs show a progress indicator with a cancellation control plus concurrency limits to avoid competing with pipeline build capacity.

GitLab Dedicated now hosts the AI Gateway for the Duo Agent Platform inside each customer's single-tenant AWS environment, keeping AI inference within the customer's AWS region and data residency boundary. It supports Amazon Bedrock as a model backend and adds bring-your-own-model configuration so customers can connect preferred model providers under their own data classification rules, unlocking all Duo Agent Platform capabilities — including automated custom code reviews, fixing failed pipelines, and security analysis — against customer-selected models on Dedicated deployments.

Polaris adds bidirectional status and triage sync across GitLab Issues, GitHub Issues, Jira, ServiceNow, and Azure Boards — triage states, ticket statuses, and fix-by dates now propagate automatically in both directions, with fix-by date changes in Polaris updating due dates in Jira, ServiceNow, and Azure Boards and vice versa. GitLab Issues, GitHub Issues, and ServiceNow become new native bug-tracking targets supporting individual and bulk export, single-ticket bundling for ServiceNow, and policy-driven issue creation. Configurable Jira field mapping lets teams map Polaris issue data to standard or custom Jira fields and customize ticket descriptions, and configurable per-project bug-tracker integration mappings can be saved as reusable templates scoped to all branches or the default branch only. Linked tickets close automatically when findings are dismissed or absent in a subsequent scan, and a reappearing issue creates a new ticket to preserve audit trails.

Polaris now generates AI-assisted fix PRs for SAST findings, triggerable CI-natively via the Bridge CLI from GitHub ActionsGitHub ActionsGitHub's built-in CI/CD system that runs automated workflows triggered by repository events, letting cyber tools automate testing, building, and releasing without managing separate pipeline infrastructure., GitLab CI, Bitbucket Pipelines, Jenkins, or Azure DevOps pipeline configuration with no server-side scheduling required. Fix PRs can also be created on-demand or in bulk from the UI, or automatically via issue policy actions, with per-branch or per-run caps on open PRs and filtering by CWECWECommon Weakness Enumeration, a catalog of recurring software and hardware weakness types. It describes classes of mistakes, such as injection, rather than individual vulnerability instances., issue type, language, and file pattern. A new View issue policies action on a SAST finding surfaces which policies, compliance requirements, and governance rules are attached to it.

Security teams can now define rule profiles at organization, application, project, or branch level with hierarchical inheritance, enforced automatically across CI pipelines, CLI scans, and integrations. The feature adds RBACRBACRole-based access control, where permissions are assigned to roles such as reader or administrator and users receive those roles. It makes access easier to manage than individual permission lists., audit trails, and version awareness, with visibility into which rule profile was used for each scan.

Polaris adds outbound-only secure tunnel connectivity (no inbound firewall changes required) to self-hosted Azure DevOps Server, GitLab Self-Managed, Bitbucket Server and Data Center, and on-premises Jira, built on Bridge CLI and Teleport infrastructure. Multitunnel support for on-premises Jira allows multiple concurrent tunnels, with tunnel setup restricted to org admins via RBAC and full audit logging of tunnel activity.

The Rapid Scan Static engine extends seven existing Java JWTJWTAn open standard (RFC 7519) for encoding claims as a signed, optionally encrypted JSON object passed between parties. Cyber tools use it to verify identity and permissions without querying a central session store. checks (algorithm confusion, nonexpiring tokens, ignored expiration/start times, missing claims validation, untrusted decoding) to Scala, adds six new Scala web security checks covering permissive and credentialed CORSCORSA browser mechanism that lets a server declare which foreign origins may read its HTTP responses, enforced by the browser to block unauthorized cross-site requests from JavaScript. origins, excessive preflight age, missing TLS, and disabled security headers across http4s and Play, adds a new crypto bits private key pattern to detect embedded cryptographic private keys in source and configuration files, and adds support for Go 1.26 and Dart 3.11.

The fAST SCA engine adds new package manager detectors for UV (Python), Rush monorepo tooling, a tree-based Conda detector, and Ant/Ivy dependency extraction. It also extends ecosystem coverage to Poetry, pnpm, SBTSBTA build tool for Scala and Java projects, maintained by the open-source community, that handles compilation, dependency resolution, and packaging using an incremental build model to speed up iterative development., SetupTools, Bazel 9, .slnx solution files for NuGetNuGetMicrosoft's package manager for .NET, hosting a public registry of reusable libraries so tools can pull in dependencies without bundling or manually maintaining third-party code., buildless go.mod analysis for Go, and Cargo workspace support with include and exclude controls.

Polaris adds webhook-triggered SAST and SCA scanning on push, branch creation, and pull/merge request activity for self-hosted SCMs, with PR/MR decoration including severity breakdowns and Code Insights reports for Bitbucket.