The Cyber Toolchain — issue 058, August 28, 2026

sslstrip now exposes sslstrip run, sslstrip doctor, sslstrip cleanup, and sslstrip --version subcommands, replacing the old monolithic entry point; the top-level sslstrip.py script is removed entirely, so any code or scripts invoking it directly will break.

$ sslstrip doctor

Adds --traffic-log and --traffic-log-file flags for unredacted traffic logging; both flags are required together to enable logging, gating it from being turned on accidentally.

$ sslstrip run --traffic-log --traffic-log-file /tmp/strip-session.log

Adds managed nftables lifecycle with an exclusive /run/sslstrip lock, atomic JSON nft operations, privilege drop, and automatic forwarding restore, replacing the iptables-only setup; the iptables-based managed path is removed and nftables is now the first-class managed firewall backend.

Adds structured HTTP error responses: missing Host header → 400, CONNECT → 405, protocol upgrades → 426, upstream/DNS/TLS failures → 502, and timeouts → 504.

Adds a shared Twisted Agent with verified TLS policy and connection pool, orchestrated via native async def and Deferred.fromCoroutine; the old Twisted HTTPClient, factory, DnsCache client stack, and global singletons are removed.

Adds a --ca-file flag to supply extra lab certificate authorities for TLS verification in controlled environments, used by the shared Twisted Agent's verified TLS policy when connecting upstream.

$ sslstrip run --ca-file /etc/lab/ca-bundle.pem

Adds native gzip, Brotli 1.2.0, and stdlib Zstandard decompression, including handling of stacked encodings, with an 8 MiB decoded-body limit.

Packaging now requires Python 3.14+ (requires-python = ">=3.14") with exact pins for Twisted[tls] and Brotli via uv_build and a locked uv.lock; environments on Python 3.x below 3.14 will not work. mypy-based type checking is replaced by strict ty, breaking mypy-based CI or type-checking workflows.

Profiles can now generate implants via external builders using profiles generate external builder, decoupling build configuration from the server.

Default BOF execution now uses Reflektor instead of the previous in-process execution model, and implant extension loading also switched to Reflektor for reflective in-memory execution. Native extension support was added on macOS and Linux, previously Windows-only, broadening cross-platform BOF/extension use.

Adds c-archive as a supported Go output format for implant generation, enabling implants to be embedded as C static libraries.

A new repeatable -header flag (format 'Name: Value') sends custom HTTP request headers with every probe, enabling scans against hosts requiring auth tokens, custom user-agents, or other header-gated content.

$ webanalyze -host internal.example.com -crawl 2 -header 'Authorization: Bearer eyJhbGc...' -header 'X-Custom-Header: value' -output json

Veracode Scan for VS Code v1.16.3 adds a Current File filter to Static Scan Findings that scopes displayed findings to the file open in the editor, an option to filter and group Static Scan Findings by file name (with findings grouped by severity within each file), and an option to change the region during OAuth authentication, with the current or most recently used region shown on the OAuth screen.

The Veracode YAML file for GitHub Workflow Integration v2.1.3 now supports custom predependency commands, enabling project preparation steps to run before a scan.

Adds an isActive parameter to the BOM upload API endpoint, allowing projects to be marked active or inactive at upload time. Adds support for compressed SBOMsSBOMsA 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. (e.g. gzip) in multipart BOM uploads, and exposes workflow run status on the BOM token polling endpoint.

$ curl -X POST 'https://dtrack.example.com/api/v1/bom' \
  -H 'X-Api-Key: <api-key>' \
  -F 'autoCreate=true' \
  -F 'isActive=true' \
  -F 'projectName=my-app' \
  -F 'projectVersion=1.0.0' \
  -F 'bom=@bom.cdx.json.gz;type=application/gzip'

Implements initial support for CISA's Known Exploited Vulnerabilities (KEV) catalog, including use of KEV status in component and vulnerability policies and inclusion of a KEV flag in notifications. Adds REST APIREST APIA web API design style where clients interact with resources over HTTP using standard methods like GET, POST, and DELETE, giving cyber tools a predictable, language-agnostic way to integrate with external services. endpoints to view and trigger KEV mirroring runs, a KEV dimension to time series metrics, and a VulnCheck KEV data source.

Forwards a group_by parameter to the DefectDojo (v5) integration and also forwards analysis.detail to DefectDojo (v5), enriching exported findings.

Adds --create-pr, which after --fix commits changes and opens a GitHub pull request automatically (requires the gh CLI); --base <branch> sets the base branch for --create-pr and defaults to main; and --debug writes verbose runtime and network diagnostics to a timestamped log file for troubleshooting scan failures.

$ cve-lite --fix --create-pr --base develop

$ cve-lite --debug

$ cve-lite --fix --create-pr --base develop

$ cve-lite --debug

Adds destroy = false to resource lifecycle blocks to prevent a resource from ever being destroyed, e.g. lifecycle { destroy = false } inside a resource block.

resource "aws_s3_bucket" "critical" {
  bucket = "my-critical-bucket"

  lifecycle {
    destroy = false
  }
}

Adds on_failure modes halt, taint, and continue to resource action triggers, adds before_destroy and after_destroy event support to Actions, adds a caller symbol in Action configurations exposing the calling resource's object value, and allows -invoke to be combined with -target to specify the calling resource instance when multiple resources trigger the same action.

Adds -format=mermaid flag to terraform graph to output the dependency graph in Mermaid format, e.g. terraform graph -format=mermaid > graph.md for embeddingembeddingLists of numbers a model produces to represent the meaning of text, images or code, so similar items sit close together. They power semantic search and RAG. in docs or wikis.

$ terraform graph -format=mermaid > graph.md

The terraform stacks command now automatically infers the target hostname from credentials.tfrc.json when neither TF_STACKS_HOSTNAME nor TF_CLOUD_HOSTNAME is set.

Starting September 28th, 2026, the default Copilot code review effort level changes from Lite to Balanced for all repositories and organizations; repos or orgs relying on Default meaning Lite must explicitly set Lite before that date. Organization-level and repository-level defaults work independently — the org default applies to repos without their own setting, while the repo default governs automatically requested reviews — and for manually requested reviews the effort level (Lite, Balanced, etc.) can be chosen directly from the Reviewers bar in the pull request UI.

Copilot Chat on github.com, Copilot Chat in GitHub Mobile, and Copilot cloud agent are merging into a single unified experience and policy, enabled by default no earlier than September 28th, 2026. As part of this shift, chat data retention extends from 28 days to the life of the account under the agent sessions model, and the cloud agent gains a Sandbox for a faster cloud execution environment. Opting out of the unified experience after September 28th, 2026 removes access to Copilot on github.com and GitHub Mobile entirely, and organizations must confirm the policy via Copilot settings > Copilot cloud agent (coming soon) before that date.

A new opt-in early-access track offers two tiers: Closed Preview (invite-only feedback cohorts) and Open Preview (available to any eligible customer, fully supported), both free. Activate it in the MyJFrog portal under Settings > Preview by enabling the Enable Preview toggle, accepting the Preview Terms of Service in the Enable JFrog Preview dialog, and then selecting specific preview features to activate on your subscription.

⤢ ⤢

New ClickUp integration (beta) lets teams create ClickUp tasks from Socket alerts via Settings → Integrations → ClickUp. Automated ticketing rules route matching findings into a chosen ClickUp Workspace and List with configurable priority (Urgent, High, Normal, Low, or Auto to map Socket priority automatically), tags, and assignee. Bi-directional status sync keeps alerts and tasks aligned — clearing an alert can move the linked task to completed, and closing or completing a task in ClickUp can triage-ignore the corresponding Socket alert — with updates delivered via Workspace webhooks. Manual task creation is available per alert via the Create Ticket button, selecting a target Workspace and List shown as Space / List or Space / Folder / List.

⤢

New Settings → Integrations → Asana UI lets an Owner or Admin connect one or more Asana workspaces via OAuth with no separate app to install. A Create Ticket button on any alert manually creates a linked Asana task, selecting workspace, project, tags, and assignee, with a bidirectional link between alert and task. Automated ticketing rules let teams select triggering alert events, add conditions (e.g. type = supply chain, priority = high), choose the target workspace and project, set tags/assignee, and configure two-way sync behavior. Two-way sync marks the Asana task complete when an alert clears, and completing the task can triage-ignore the linked Socket alert.

⤢

Adds continuous analysis of extensions published through Microsoft Edge Add-ons, covering malware, credential and clipboard theft, data exfiltrationdata exfiltrationMoving data out of a system or organization without authorization. It can happen through network transfers, removable media or other channels., remote code loading, obfuscation, and impersonation — the same detections previously available for Chrome and Firefox. Enables version-change comparison, surfacing meaningful differences in code, permissions, network activity, and behavior between releases under the same extension identity. Identifies command-and-control infrastructure, suspicious endpoints, remote payload sources, and changes in contacted domains, and connects related extensions across stores through reused code, shared infrastructure, publisher patterns, and coordinated campaigns. This pairs with existing Chrome Web Store analysis to close the cross-store coverage gap.

⤢

Hunter Agent is now GA on SonarQube Cloud (Enterprise plan), an AI security agent that detects broken access control (IDORIDORInsecure Direct Object Reference, a web vulnerability where an application exposes internal identifiers (IDs, filenames) without access checks, letting an attacker substitute another user's ID to reach their data., missing authorization checks, privilege escalationprivilege escalationGaining permissions beyond those an account or process initially had. It can turn limited access into administrator control or access to another user's resources., CSRFCSRFCross-site request forgery, an attack that makes a signed-in browser send an unwanted request to a site. The site may accept the request using the victim's existing session.), business logic flaws (skipped workflow steps, missing rate limits), and authentication/session management flaws (session fixation, non-expiring sessions, weak password recovery, missing MFAMFAMulti-factor authentication, requiring proof from different categories, such as a password and a hardware key, to sign in. It reduces reliance on a password alone.) — vulnerability classes SAST structurally cannot evaluate. Every candidate finding is validated for exploitability via a dedicated subagent before surfacing, achieving 80–90% average precision, and confirmed findings appear directly as SonarQube issues in the existing triage workflow with no separate portal, dashboard, or installation required. It supports on-demand triggering or scheduled runs (weekly or monthly) against a main branch, delivering findings asynchronously once confirmed; scales to codebases beyond 10 million lines of code via a structured multi-phase pipeline (Analyze, Explore, Validate, Synthesize) that cuts re-scan time on large repositories by roughly half through incremental scanning; works across the 40+ programming languages already supported by SonarQube with no new configuration beyond pointing it at an existing SonarQube project; and every finding carries an auditable discovery path showing how the agent reached its conclusion.

A new Credentials page in the StepSecurity dashboard aggregates credential source coverage fleet-wide, showing per-source device count, plaintext vs. protected breakdown, location count, and entry count, filterable by category (Cloud, Source control, Package registries, Containers, Infrastructure) and exportable as CSV. It covers thirteen credential source types across five categories — ~/.npmrc, ~/.pypirc, ~/.vault-token, ~/.netrc, ~/.aws/credentials, ~/.aws/config, ~/.docker/config.json, gcloud application default credentials, kubeconfig, ~/.ssh private key files (id_rsa, id_ed25519, id_ecdsa), and ~/.git-credentials — with a per-source detail view listing affected devices labelled Plaintext, Protected, or both and direct links to each machine. On macOS, Full Disk Access must be granted to the agent via MDMMDMMobile device management, centrally configuring and managing enrolled devices. Administrators use it to apply settings, distribute applications and enforce device security policies. to reach TCC-protected credential paths for complete results. Available from agent v1.16.0; existing deployments gain the page automatically after each device's next scan.

⤢ ⤢ ⤢

A new Browser Extensions page in the StepSecurity dashboard shows a fleet-wide table of every detected extension — name, extension ID, browser, store, version, device count, and last-seen timestamp — filterable by browser, capability, store, and engine, with an Export CSV download. Raw browser permissions are rolled into ten named capabilities (Global website access, Global request observation, Script injection, Global script injection, Cookie access, Global cookie access, Browsing history, Debugger access, Clipboard read, Extension management) reflecting what the browser currently honours rather than just what the extension's manifest declares. A per-device detail panel lists affected devices grouped by version (first 500 devices shown) for rapid scoping against version-specific advisories. Covers Chrome, Edge, and Firefox on macOS, Windows, and Linux — including Firefox under snap and Flatpak and Edge under Flatpak — available from agent v1.16.0 with no changes required for existing deployments.

⤢ ⤢ ⤢ ⤢

Aikido's agents install a real APK on a test device, log in as a user, and attack the app over ADB, with every finding exploit-confirmed before it appears in the report and reproduction steps including the exact ADB commands needed to replay the exploit. Coverage includes unsafe WebView usage, deep-link and exported-component abuse, client-side authentication issues, injection flaws, broken access control, weak session handling, unsafe API behavior, IDORs, cross-tenant access, and authentication bypasses. Assessments are rightsized by Aikido's indexer to the actual app size rather than charged a flat fee, and each run produces a recon report and attack plan with diagrams, threat scenarios, and the full agent reasoning trace (including vulnerabilities attempted but not confirmed). The capability is currently white-box only, requiring source repository access alongside the APK (black-box support is coming soon), and the uploaded build must have certificate pinning, root detection, emulator detection, and RASP disabled.

⤢ ⤢ ⤢ ⤢

After merging an AutoFix pull request for a confirmed Android finding, users can retest against the patched build without rescoping: open the original Android assessment in the Aikido console, go to the overview tab, upload the rebuilt APK, then click Retest on any finding from the initial pentest.

Adds a --supplier flag to the Manifest CLI for associating an uploaded SBOM or binary directly with a named supplier at publish time.

New supplier portfolio view ranks every third-party vendor by risk score, asset count, open vulnerabilities, and last-updated date in a single screen. A companion supplier detail view surfaces a vendor's full footprint — every asset, version, risk score, component count, and vulnerability count — in one click. Aggregated supplier risk scoring rolls up asset-level risk into a single per-supplier score for direct vendor-to-vendor comparison, and prioritized supplier findings stack-rank vulnerabilities by severity and product impact across a supplier's portfolio. The vulnerability and component pages were also extended to surface third-party products alongside first-party assets, with clear visual distinction between what teams build and what they buy.

Introduces binary-generated SBOMs for vendors who cannot provide one, covering C++, unmanaged code, and embedded systems where traditional 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. falls short, with results folded into that supplier's record.

Introduces clearly-cached, a public caching proxy at clearly-cached.sbomify.com that sits in front of ClearlyDefined and returns a flat 0.4KB projection of license fields, resolving 76 of 76 packages in 5s warm vs 31 of 76 in 474s direct. SBOMIFY_CLEARLY_CACHED_URL env var points the action at a self-hosted or local instance instead of the default.

- uses: sbomify/sbomify-action@v26.8.0
  env:
    SBOMIFY_TOKEN: ${{ secrets.SBOMIFY_TOKEN }}
    SBOMIFY_CLEARLY_CACHED_URL: 'http://clearly-cached.internal:8080'

New Security Advisories module lets you author vendor advisories with CVSSCVSSCommon Vulnerability Scoring System, a method for describing a vulnerability's characteristics and severity. Its score is not the probability that attackers will exploit it. vector, affected/fixed version ranges, product associations, and a timestamped timeline, published directly to the Trust Center. Advisories support three visibility states — Private, Public, and Gated — where Gated requires both a workspace-level NDA grant and a link to a product the reader can already access; per-advisory product filtering redacts names the reader is not entitled to see.

Adds SBOMIFY_FETCH_RUNTIMES=0 env var to opt out of on-demand tool fetching for air-gapped environments; per-ecosystem tool bundles are now fetched from sbomify/sbom-tools at runtime, pinned by digest, verified, and attested, replacing tools baked into the image, dropping the container image from 515MB to 106MB. The generator fallback chain is now strict: a generator that claims a project and then fails aborts the run instead of silently falling through to a lower-quality scanner.

- uses: sbomify/sbomify-action@v26.8.0
  env:
    SBOMIFY_TOKEN: ${{ secrets.SBOMIFY_TOKEN }}
    SBOMIFY_FETCH_RUNTIMES: '0'

New CRA patch SLA targets, defaulting to Critical 7 / High 30 / Medium 90 days, gate five computable KPIs: mean time to triage, mean time to remediate, open criticals with KEVKEVCISA's Known Exploited Vulnerabilities catalog, a list of vulnerabilities with evidence of exploitation in real attacks. It helps teams prioritize fixes for problems attackers already use. split, patch SLA compliance, and VEXVEXA machine-readable statement from a vendor saying whether a CVE in a component actually affects their product. It is how a scanner is told to stop flagging a vulnerability that is not reachable. publication latency.

New risk_accepted finding state requires a business justification, named risk owner, and expiry date; it publishes downstream as exploitable with a will_not_fix response in CycloneDXCycloneDXAn open standard for describing software components and their dependencies, vulnerabilities and other supply-chain information. Tools use it to exchange bills of materials and related evidence. rather than not_affected, to avoid false claims.

Four new CSV exports covering inventory, licenses, findings, and vulnerabilities, scoped to workspace, product, release, or component, reading the same latest-per-component data and live VEX suppressions as the dashboard.

Adds first-class lock-file and manifest support for Gradle (gradle.lockfile), Swift (Package.swift), PHP (composer.json via a proper Composer bundle), Rust crates with only Cargo.toml, .NET project files, Haskell, Erlang, and Clojure. Manifests listing version ranges now read the adjacent lock file to record the versions actually shipped.

security.txt generation follows BSI TR-03183-3 shape with separate CSIRT contact and report-page slots, an optional CSAFCSAFAn OASIS open standard for distributing machine-readable vulnerability advisories, letting cyber tools automate ingestion and correlation of vendor-issued security disclosures at scale. pointer, and an Expires field clamped to at most one year; it shares data with the CRA wizard.

CRA wizard now captures EU Authorized Representative details and mandate per Article 22/25, rendering them on the Declaration of Conformity; missing mandatory fields are surfaced loudly rather than silently omitted.

Trust Center redesigned with a filter sidebar, dense advisory table with per-product version rows rendering version ranges as comparison strings (e.g. >= 11.2.0, < 11.2.4-h20), live faceting, and a JavaScript-free plain form fallback.

End-of-life readiness check surfaces unpatched critical and high findings, drafts an EOL announcement through the advisory channel, and pins the final SBOM and VEX to a versioned release instead of a floating latest pointer.

Malicious packages are now a distinct structural category, reclassified at read time across all historical scans without rescanning, and no longer fall through to a default medium severity.

Adds delete protection for KMS keys, preventing accidental key removal. Enable it in the Infisical console under project > Key Management, select a key, open its settings, and enable Delete Protection.

The infisical-nkp Helm chart now requires config.siteUrl to be set in its configuration.

Adds --with-openssl=DIR configure flag for builds where OpenSSL is installed in a non-standard path.

Adds mTLS support, enabling client certificates when testing targets that require mutual TLSmutual TLSMutual TLS, an encrypted connection where both client and server present certificates to prove their identity. It helps services verify who they are communicating with. authentication, and custom request header injection, allowing arbitrary headers to be sent with slow HTTP test payloads.

The DataDome API reference gains several new fields for detection requests: AcceptLanguage for passing the request's Accept-Language header to the detection engine, Via for forwarding proxy chain information, SecCHUAFullVersionList for passing the Sec-CH-UA-Full-Version-List client hint header, and CustomFieldString1, CustomFieldString2, and CustomFieldString3 for sending arbitrary custom string data alongside detection requests.

Adds Threema account, contacts, and messages parsing, requiring a new sqlcipher3 version 0.6.2 dependency to decrypt Threema's encrypted databases.

Adds Truecaller call history, contacts, SMSSMSShort Message Service, the protocol built into cellular networks for sending text messages between phones; cyber tools use it for alerting, two-factor authentication delivery, or phishing simulation over mobile channels. senders, call cache, and settings parsed from tc.db, insights.db, and calling-cache.db.

Adds Ryanair boarding pass parsing with decryption using the key stored in the app's own preferences, plus bookings, flights, passengers, and sessions.

Adds Uber location artifacts with KMLKMLAn XML-based file format maintained by the OGC for storing geographic data such as points, paths, and polygons. Cyber tools use it to visualize location-based intelligence on maps. output for coordinates and timestamps, plus sessions, payment profiles, telemetry messages, cached images, and app settings.

Gmail parsing now reads every account store across all Android users instead of only the first, and Gmail bodies are now shown as readable text with a dedicated Links column.

Adds parsers for over 35 new Android apps: Uber, Facebook, Threads, Twitch, Zoom, PayPal, Revolut, Outlook, Yahoo Mail, Adobe Acrobat Reader, Disney+, Amazon Prime Video, Pinterest, Hinge, NordVPN, Roblox, eBay, Etsy, Vinted, Booking.com, Ryanair, DHL, CapCut, Shazam, Truecaller, SimpleX Chat, TeleGuard, Threema, Opera Browser, Samsung Bixby Routines, Spotify (6 artifacts), and Private Photo Vault.

A new admin/scripts/check_artifact_paths.py CI guard, with --self-test and --list flags, fails when an artifact's paths tuple contains two patterns differing only in case; it is wired into the lint workflow.

iLEAPP now reports real device/source paths instead of the examiner's local folder structure, a change enforced across 83+ modules including Biome, CNACNACVE Numbering Authority — an organization authorized by MITRE to assign CVE IDs within a defined scope, letting cyber tools reference vulnerabilities under a globally consistent identifier., and Privacy Accounting artifacts.

The installer can now create a separate bootable USB drive preloaded with a BitLocker TPM exploit, supporting YellowKey and GreatXML, turning what was previously a manual multi-step process into a guided setup click.

Configuration files nested below cfg/<vendor>/ are now correctly installed during package setup; the Linux kernel driver is now loaded via modprobe before searching for the driver file, improving portability of installed builds; and the manual is now installed to a path relative to the install prefix rather than a hardcoded location.

Adds an is_enabled field to controls, enabling filter_enabled filtering in object lists of controls.

Control Center is a no-cost capability of Inbound Email Security offering two detection layers. Custom AI Models (Generally Available) let analysts describe a detection pattern in natural language to create a dedicated model, which is backtested against recent organizational mail and can run in Passive mode before full enforcement. Custom Rules (Early Access) let teams build organization-specific enforcement using more than 50 message attributes combined with AND, OR, and Except When logic — no proprietary query language required — with per-user or org-wide scoping, and can also be validated in Passive mode before going Active. Each Custom Rule and Custom AI Model assigns a verdict of Attack, Spam, or Safe, with Custom Rules taking precedence over Custom AI Models, which in turn override Abnormal's core AI detection; decision transparency shows which of the three layers made each detection decision.

Email DLP Rules (Early Access) let security and compliance teams define, validate, and enforce outbound email policies via a rule builder supporting regex, phrase matching, metadata conditions, and Boolean logic — including AND/OR nesting, condition groups, and exclusions — with prebuilt templates for data types like financial information and PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations.. Rules scan outbound content and supported attachments and can be scoped by department or workflow. The AI Triage Agent evaluates each flagged message against the rule's stated intent, sender-recipient relationship, and surrounding context to distinguish genuine violations from routine business traffic, auto-releasing benign messages and quarantining likely violations; it also generates a plain-language description of each rule's intent and editable exclusions at creation time, and rules can be validated against sample emails before live deployment. Every Triage Agent verdict is recorded in a new Outbound Log for audit visibility.

AI Phishing Coach gained: Precision Training, which automatically adjusts simulation frequency and difficulty per employee based on behavior signals and response history, giving more frequent/harder simulations to elevated-risk employees and fewer to low-risk ones, plus reinforcement training when needed; Attack Creator, which generates a realistic simulation from a plain-language scenario description and stores it in the Attack Library; Campaign Creator, which schedules and sends ad hoc or baseline simulation campaigns from the Attack Library, including AI-generated attacks; Training Creator, which generates custom training video modules from a prompt and imagery, with module recommendations driven by Abnormal's observed threat patterns; QR Code Phishing Simulations, which train employees to treat QR codes with the same suspicion as suspicious URLs and render even in email clients that block images; and In-Thread Follow-Up, which re-engages employees who neither clicked nor reported a simulation by sending a follow-up inside the same email thread, mirroring real attacker tactics.

The new --not-a-security-boundary flag makes failures in sandbox setup steps (such as remounting a submount) non-fatal, enabling fail-open behavior for use cases that do not require a strict security boundary.

$ bwrap --not-a-security-boundary --ro-bind /usr /usr --proc /proc --dev /dev --unshare-pid bash

A new assume_kernel meson build option lets builders specify a minimum kernel version; specifying 5.6.0 or later disables the fallback implementation of openat2(RESOLVE_IN_ROOT) and removes backward compatibility code for older kernels, resulting in hard failures at runtime on unsupported kernels.

Adds CreateSandboxRequest.lifecycle (preStart / periodic hook types per OSEP-0020), transported to Kubernetes workloads via the reserved OPEN_SANDBOX_LIFECYCLE env var; the Docker provider explicitly rejects lifecycle requests. Config persists at $HOME/.execd/lifecycle.toml with EXECD_LIFECYCLE_CONFIG as an exact override path; execd starts its HTTP server before preStart and only releases the user process after preStart succeeds, with bootstrap sync and preStart shutdown fail-closed under TERM/KILL watchdogs. The preStart.timeoutSeconds server and OpenAPIOpenAPIA vendor-neutral specification for describing REST APIs in a machine-readable format, maintained by the OpenAPI Initiative. Cyber tools use it to auto-generate clients, documentation, and validation from a single source of truth. limit was raised from 300s to 10800s (3 hours); the periodic limit stays at 300s.

$ EXECD_LIFECYCLE_CONFIG=/etc/sandbox/lifecycle.toml execd

Adds kubernetes.pool_acquisition_timeout_seconds config key (default 30s) so the server waits a bounded window on pool exhaustion, then returns HTTP 429 with KUBERNETES::POOL_CAPACITY_EXHAUSTED and Retry-After: 5 instead of a generic 504. Pool exhaustion is now machine-identifiable via a controller-published PoolAllocationPending / PoolCapacityExhausted condition, and getSandbox surfaces an optional Sandbox.allocation block (mode: pool, poolRef, state: allocated) for confirmed Kubernetes pool allocations.

[kubernetes]
pool_acquisition_timeout_seconds = 60

[egress]
timeout_seconds = 60

Enables execd to run as PID 1 via EXECD_INIT=1 (or server-side runtime.execd_run_as_init), providing a single reaper with signal forwarding (TERM/HUP/USR1/USR2/WINCH), entrypoint-owned exit-code propagation, and a subreaper fallback for the Pool path; classic topology is unchanged and this is off by default.

$ docker run -e EXECD_INIT=1 opensandbox/execd:v1.1.0

$ docker run -e EXECD_INIT=1 opensandbox/execd:v1.1.0

Adds a new opensandbox-launcher pre-exec hardening pipeline (env strip → KEEPCAPS → bounding-set trim → no_new_privsno_new_privsA Linux process flag, set via `prctl()`, that prevents a process or its children from gaining elevated privileges through `setuid` binaries or file capabilities, limiting exploit escalation paths. → identity drop → seccomp → execve), with every layer degrading fail-open. A new [landlock] layer applies a filesystem allowlist, failing open below ABI/kernel 5.13 and reporting unsupported state. Per-layer state is reported in a new hardening object on GET /v1/isolated/capabilities.

$ curl -s GET https://<host>/v1/isolated/capabilities | jq '.hardening'

$ curl -s https://<sandbox-host>/v1/isolated/capabilities | jq '.hardening'

Accepts OPENSANDBOX_SECURE_ACCESS_KEYS and OPENSANDBOX_SECURE_ACCESS_ACTIVE_KEY env overrides for secure-access signing keys; the Helm chart adds server.gateway.secureAccess.existingSecret (mutually exclusive with plaintext keys) so GitOpsGitOpsA practice where Git repositories serve as the single source of truth for infrastructure and application state, with automated systems continuously reconciling live environments to match what is declared in version control. flows never commit key material.

server:
  gateway:
    secureAccess:
      existingSecret: opensandbox-signing-keys

Adds opt-in QEMU VMState pause/resume for runc-based sandboxes: capture of guest memory, vCPU, and device state via QMP migration, stored as separate OCIOCIOpen Container Initiative standard, maintained by the Linux Foundation, defining the image format and runtime spec that container tools share so images built once run anywhere OCI-compliant. images keyed by immutable SHA-256SHA-256A cryptographic hash function that turns data into a fixed 256-bit fingerprint. It is used to check integrity; hashing does not encrypt the original data. digests; supports pooled BatchSandbox pause, detachment, and standalone restoration.

New [docker] config fields sandbox_env and sandbox_binds apply fleet-wide environment variables and host bind mounts to every Docker sandbox container; request env wins over sandbox_env, and config binds precede request-derived binds.

[docker]
sandbox_env = { "LOG_LEVEL" = "info", "REGION" = "us-east-1" }
sandbox_binds = ["/etc/ssl/certs:/etc/ssl/certs:ro"]

Diagnostics API now honors scopes: logs: container|all and events: runtime|lifecycle|all, returning 400 DIAGNOSTICS_SCOPE_UNSUPPORTED for unknown scopes; deprecated no-scope responses still work but now include a Deprecation header.

Adds generic HTTP request metrics via OTLPOTLPOpenTelemetry Protocol, a standard way to send traces, metrics and logs between applications, collectors and monitoring systems. It lets telemetry move across compatible tools.: a thin ASGIASGIA Python async web server interface specification that defines how servers and applications exchange HTTP and WebSocket messages; cyber tools use it to serve high-concurrency APIs without blocking on I/O. middleware exports server.http.request.duration (ms) with http_method, http_route, and http_status_code attributes; no-op when [otel] is disabled.

New [egress] timeout_seconds config key (default 30s; non-positive values rejected) lets Docker deployments tune how long the server waits for the sidecar health endpoint. An experimental OPENSANDBOX_EGRESS_MITMPROXY_EXTRA_PORTS env var appends extra TCPTCPA network protocol that delivers a reliable, ordered stream of bytes between two endpoints. It detects lost data and retransmits it. ports to the always-on 80/443 interception list for the egress transparent MITM, capped at 15 ports.

[kubernetes]
pool_acquisition_timeout_seconds = 60

[egress]
timeout_seconds = 60