The Cyber Toolchain — issue 044, August 13, 2026

Adds ServiceNow to the Mappings Manager: define how Dradis fields map to ServiceNow fields, with Liquid templating support, across the Application, Container, and Vulnerability Response modules, so mapped fields auto-populate whenever a Vulnerable Item is created. Configure via Mappings Manager → ServiceNow → New ServiceNow Mapping.

Adds an HTML / PDF Gateway export option directly from a project's Export tab, generating themed HTML output (including the Athena theme, recommended for print-to-PDF) without requiring access to the Results Portal, alongside the existing Word, Excel, and CSV export options.

Adds multiplayer Echo chat on Issues via the Echo tab: start a conversation from a saved prompt, send follow-up messages, and watch Roslin's replies stream live to all project members viewing the same Issue.

Adds CSV import support to the Mappings Manager: column-to-Issue/Evidence field mappings are defined once per Report Template and applied automatically to future uploads that use the same column headers.

Unit 42 Frontier AI Exposure Analysis now deploys GPT-5.6 Daybreak — previously unavailable for commercial use — directly in customer environments to find exposures, test exploitability, validate attack pathsattack pathsA possible sequence of steps an attacker could take to reach a valuable target. It connects weaknesses, permissions and reachable systems rather than treating each finding separately., and prioritize remediation.

The Prisma AIRS Runtime API now integrates natively with OpenAI Codex Enterprise: pasting the Prisma AIRS API key and endpoint into the Codex Global Admin Console routes all user prompts across an organization through real-time inspection with no client-side agents required. This also extends Enterprise DLP controls to Codex, scanning developer inputs — including logs, config snippets, and code functions — for proprietary data and credentials before they reach the AI model.

A new Jira integration class (in prowler/lib/outputs) exposes send_findings(findings, project_key, issue_type), test_connection(redirect_uri, client_id, client_secret, raise_on_exception), get_projects(), get_available_issue_types(project_key), get_access_token(), refresh_access_token(), auth_code_url(), and input_authorization_code() to push findings into Jira Cloud via the Atlassian 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.. The OAuth 2.0OAuth 2.0An open authorization framework (IETF RFC 6749) that lets a service grant limited API access to a third party without sharing credentials, giving cyber tools a standard way to authenticate against external platforms. authflow uses AUTH_URL (https://auth.atlassian.com/authorize), TOKEN_URL (https://auth.atlassian.com/oauth/token), and API_TOKEN_URL (https://api.atlassian.com/oauth/token/accessible-resources) with scopes read:jira-user, read:jira-work, and write:jira-work; a developer guide documents building further custom integrations (e.g. Slack, AWS Security Hub) in the same prowler/lib/outputs folder.

from prowler.lib.outputs.jira import Jira

jira = Jira(
    redirect_uri="http://localhost:8080",
    client_id="client_id",
    client_secret="client_secret",
)
jira.send_findings(findings=findings, project_key="SEC", issue_type="Bug")

from prowler.lib.outputs.jira import Jira

conn = Jira.test_connection(
    redirect_uri="http://localhost:8080",
    client_id="client_id",
    client_secret="client_secret",
    raise_on_exception=False,
)
print(conn.is_connected, conn.error)

The iam_policy_allows_privilege_escalation check gains 22 additional escalation combinations, and 20 new AWS privilege-escalation Attack Paths queries cover iam:PassRole paths across AWS Batch, Braket, Cognito Identity, ECSECSElastic Common Schema, an open specification maintained by Elastic that defines a standard field naming convention for log and event data, letting cyber tools share and correlate records without per-source field mapping., EMR, EMR Serverless, GameLift, Glue, EC2EC2Amazon's cloud virtual-machine service, letting tools run on-demand servers without managing physical hardware. Image Builder, Kinesis Analytics, HealthOmics, EventBridgeEventBridgeAWS's managed event bus service that routes events between cloud services, SaaS apps, and custom code using rules and targets, letting tools react to state changes without polling or managing message infrastructure. Scheduler, Systems Manager, and Step Functions, plus existing-resource abuse, permissions-boundary removal, role assumption, and IAMIAMIdentity and Access Management — a framework of policies and services for controlling who can authenticate to systems and what resources they are authorized to use, letting cyber tools enforce least-privilege and audit access. Identity Center permission-set policy injection. Each query now exposes an outcome category (code execution, 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., public exposure, resource inventory), and the Attack Paths graph was redesigned to group same-class resources into expandable nodes with counts and terminate every path in an explicit outcome node naming the destination impact.

The prowler_send_findings_to_jira 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. tool gains Cloud-only grouped dispatch: it selects failed findings by check IDs against the latest completed scan and sends one Jira work item per check listing up to 50 affected resources, with per-group error reporting.

GET /api/v1/users/me membership relationships now identify the active tenant with meta.active for both 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. and API key authentication.

Prowler API, UI, SDK, and MCP container images (and Prowler Cloud/Private Cloud production images) now publish per-architecture 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. and build-provenanceprovenanceEvidence recording where a software artifact came from and how it was built, including its source and build process. It helps a consumer check that the artifact matches a trusted build. attestations. SDK and API container builds also verify checksums of downloaded PowerShell, Trivy, and zizmor binaries before installation, and Grype now complements Trivy across container-image security gates, detecting components and vulnerabilities manifest-based scanners can miss and blocking fixable high and critical findings.

Adds batch_job_definition_no_secrets, scanning Batch job definition environment variables and command parameters for hardcoded secrets, and awslambda_layer_no_secrets_in_content, scanning Lambda layer package content for hardcoded secrets.

Adds 12 new checks for 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. Microsoft 365 Foundations Benchmark v7.0.0 across Admin Center, Defender, Entra IDEntra IDMicrosoft's cloud-based identity and access management service, formerly Azure Active Directory, used by cyber tools to authenticate users and enforce access policies in Microsoft 365 and Azure environments., Exchange Online, and Microsoft Teams, followed by 20 new Microsoft 365 Entra ID checks for the same benchmark covering password protection, default user permissions, Conditional AccessConditional AccessA policy enforcement framework, most commonly Microsoft's in Entra ID, that grants or blocks resource access based on signals like user identity, device state, and location — letting cyber tools trigger step-up auth or deny risky sessions., sessions and authentication methods, and PIMPIMProtocol Independent Multicast, an IETF routing protocol that distributes IP multicast traffic across networks without depending on a specific unicast routing protocol, letting tools monitor or manipulate multicast group traffic. and access reviews.

Adds CMMC 2.0 as a universal compliance framework covering all 149 requirements across Levels 1, 2, and 3, mapped to checks across AWS, Azure, GCP, Alibaba Cloud, Oracle Cloud, and Microsoft 365.

A guided wizard discovers Azure subscriptions within Management Groups and selectively onboards them through a single workflow. In Prowler Cloud/Private Cloud, entering the Microsoft Entra tenant ID and authenticating with a single tenant-wide service principal onboards all subscriptions under that tenant at once.

⤢ ⤢

Adds a Lighthouse AI Skills menu on every finding in Prowler Cloud, offering Contextual Fix, Triage Decision, Systemic Scope, and a free-form prompt, with follow-up actions including creating a Jira issue or muting the finding (Cloud/Private Cloud only). A new Manage Lighthouse AI role permission in role forms and role details lets this configuration permission be granted or restricted independently of other permissions.

⤢ ⤢

Google SecOps adds a public preview Cloud Logging integration (requires a Bring Your Own Project (BYOP) Google Cloud project) to monitor, debug, and troubleshoot SIEMSIEMSecurity information and event management, collecting logs from many systems and correlating events to find suspicious activity. Analysts use it to investigate alerts and incidents. ingestion pipelines and feeds via Logs Explorer, supporting both push- and pull-based ingestion mechanisms. A new Debug with logs option on the Feed management page opens Logs Explorer pre-filtered for a specific feed for fast per-feed diagnosis. Exclusion filters can be configured in the Log Router to exclude specific logs (e.g., Storage Transfer Service (STSSTSAWS Security Token Service, an Amazon Web Services API that issues short-lived credentials for IAM roles, letting cyber tools authenticate to AWS without storing long-term secrets.) logs) from being routed to Cloud Logging, and Gemini Cloud Assist can be used to investigate logging and metrics telemetry directly from the Google SecOps console.

▶ ⤢

▶ ⤢

The rich-text editor used in Cases Wall, the Use Case Upload dialog, the Report Template dialog, and the Dashboard Editor widget has been upgraded with semantic font sizes (Small, Normal, Large, Huge) and simplified table insertion/removal. As a result, the new editor no longer supports formatting inside table cells, and the Cut, Copy, and Paste buttons have been removed from the toolbar, though standard OS keyboard shortcuts still work.

A 280-line dependency-free bash hook collector script (log-tool-calls.sh), published in the elastic/elasticsearch-labs repository with a PowerShell port for Windows, wires into Cursor's hooks.json (covering sessionStart, beforeShellExecution, afterShellExecution, beforeMCPExecution, afterMCPExecution, postToolUse, postToolUseFailure, afterFileEdit, beforeReadFile, subagentStart, subagentStop, sessionEnd, stop) to record every Cursor AI agent tool call as JSONLJSONLA text format where each line is a self-contained, valid JSON object, making it easy to stream, append, and process large datasets line by line without parsing a full JSON array into memory. to a date-rotated log file readable only by the owner (chmod 0600). It covers both the IDE and the headless cursor-agent CLI (used in CI and scripted runs) through the same hooks.json, firing identical hook events, and enforces local retention of 30 days with opportunistic deletion of older files while Elasticsearch serves as the system of record.

$ sudo mkdir -p /usr/local/share/ai-hooks
sudo cp log-tool-calls.sh /usr/local/share/ai-hooks/log-tool-calls.sh
sudo chmod +x /usr/local/share/ai-hooks/log-tool-calls.sh
sudo tee "/Library/Application Support/Cursor/hooks.json" <<'EOF'
{
  "version": 1,
  "hooks": {
    "sessionStart":          [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "beforeShellExecution":  [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "afterShellExecution":   [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "beforeMCPExecution":    [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "afterMCPExecution":     [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "postToolUse":           [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "postToolUseFailure":    [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "afterFileEdit":         [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "beforeReadFile":        [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "subagentStart":         [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "subagentStop":          [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "sessionEnd":            [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }],
    "stop":                  [{ "command": "/usr/local/share/ai-hooks/log-tool-calls.sh" }]
  }
}
EOF

$ cursor-agent --print "say hello"
tail -1 ~/.config/ai-hooks/logs/tool-calls-$(date -u +%F).jsonl

Each event is tagged with a detected agent and ide value (cursor, cursor-cli, or remote) by inspecting inherited environment variables CURSOR_VERSION, CURSOR_CODE_REMOTE, VSCODE_PID, VSCODE_CWD, and VSCODE_IPC_HOOK, so IDE and CLI sessions are distinguishable within a single log stream.

$ cursor-agent --print "say hello"
tail -1 ~/.config/ai-hooks/logs/tool-calls-$(date -u +%F).jsonl

Every log entry promotes top-level structured fields — hook_event_name, tool_name, command, file_path, mcp_server, model, session_id, duration, duration_ms — plus an ECS-conformant event object (event.category, event.type, event.outcome) so entries correlate with existing Elastic security indices.

The collector can be delivered to devices outside MDM via Cursor's Cloud Distribution console, which pushes a self-bootstrapping hook command carrying the collector script gzipped and base64-encoded, and verifies it against a pinned 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. hash on every invocation.

An Elastic Agent filestream integration configuration parses the JSONL logs into structured fields, enabling ES|QL queries across all agent activity — for example, identifying which hosts ran an agent that read a .pem file in a given time window.

Zscaler Zero Trust Gateway (ZTGW) integrates with Google Cloud's Network Security Integration (NSI) to transparently steer workload traffic to Zscaler for deep inspection without application changes, manual tunnels, or route table modifications, and is deployable in under 10 minutes. It supports Single 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., Multiple VPC, and Shared VPC Google Cloud topologies, uses Google Cloud Security Profiles and Profile Groups to control traffic steering and enable policy-driven enforcement, and provides unified visibility across GCP, other clouds, and on-premises environments through the Zscaler Admin Portal.

Auth0's Flexible Password Policy is now generally available, replacing legacy password fields (passwordPolicy, password_complexity_options, password_history, password_no_personal_info, password_dictionary) with a single options.password_options object on database connections. It adds composition controls (minimum length, required character types, blocking of sequential or repeated characters), password history enforcement against a configurable number of previous passwords, dictionary blocking via a 10,000- or 100,000-word built-in list plus custom entries, and profile-data blocking that rejects passwords containing values from a configurable list of user profile fields. New database connections created as of July 2026 default to this policy with a 15-character minimum length and no required character types, aligning with NIST guidance, and can be created via Dashboard, Management API, Deploy CLI, Terraform, or an Auth0 SDK.

PATCH /api/v2/connections/{id} now returns 400 invalid_body if a request contains both password_options and any legacy password field (passwordPolicy, password_complexity_options, password_history, password_no_personal_info, password_dictionary) simultaneously. Automation that reads a connection and writes it back with legacy fields appended will break and must be updated to send only password_options.

Adds the /voice command (also triggerable via Ctrl+O or by holding Space) for on-device speech-to-text prompt dictation using Whisper, with no audio leaving the machine and no cloud API key required. A --continuous flag enables hands-free back-and-forth mode that auto-starts a new recording each turn.

$ /voice --continuous

Ctrl+X at spec phase checkpoints opens a spec review screen where you can read phase documents and stage line comments; these staged comments are batched into a single revision request when the checkpoint question is answered.

AGENTS.md files now load as steering context from anywhere in the workspace tree, not just the workspace root or ~/.kiro/steering/, enabling co-location of steering instructions with the code they describe.

Adds RecoverTurnRequest and CodexThread::recover_turn_if_idle to resume an interrupted regular turn with its existing turn ID and updated thread settings, including Plan-mode recovery without injecting an empty user message. Turn recovery is rejected while another turn is already active, without applying settings or queuing input.

Materializes inline visualization viewer documents in a dedicated cache under CODEX_HOME, keyed by source and artifact thread IDs, instead of alongside visualization artifacts, preventing sandboxed sessions from modifying viewer documents before they open in a browser. Visualization links are now created only when the active filesystem policy cannot write to the viewer cache, disabling links for full-disk-write sessions and applying the policy consistently when loading transcript history.

Managed Hub daemons now upgrade directionally — when another Cline install ships a newer Hub build, hosts attach to the newer daemon and prompt to update and restart (via onHubUpdateRestart) instead of two installs repeatedly retiring each other's daemons; yolo and sandbox sessions are never interrupted by this prompt. After the user accepts the dialog, the CLI self-updates automatically, running cline update under the hood and printing instructions to reconnect.

The Claude Code provider now declares its own native tools, anchors the session on the workspace directory, and loads ~/.claude plus project settings so user-configured permission rules apply; file edits under the workspace are auto-approved while command execution remains gated by Claude settings.

Every tool call now gets its own row in the session transcript with its own icon, status indicator, and expandable detail — commands render like a terminal (e.g. $ bun test) with captured output on expand, and edits show diffs inline per hunk. Running tool rows are highlighted in brand violet and settle to gray on completion, with error rows staying red. File diffs in both chat rows and the diff panel now render through a shared syntax-highlighted renderer that follows the app theme, and the thinking indicator stays visible during quiet stretches of a turn (such as while tool arguments are streaming) so turns no longer appear frozen.

Vertex AI provider adds Fable 5 and support for custom model IDs, and the global-region picker no longer hides models from the live catalog.

Compaction now respects the Max Output Tokens setting instead of a hardcoded 1024-token cap, enabling reasoning models to produce summaries instead of skipping compaction.

Token telemetry now reports disjoint per-request buckets — uncached input, cache reads, and cache writes — eliminating ~5x inflation of per-task sums on cache-heavy sessions.

Built-in TerminalTool, FileEditorTool, and TaskTrackerTool are importable from openhands.tools.terminal, openhands.tools.file_editor, and openhands.tools.task_tracker. The Agent class composes an agent from an llm and a list of Tool objects, and the Conversation class manages the interaction lifecycle via send_message() and run() within a given workspace path.

import os
from openhands.sdk import LLM, Agent, Conversation, Tool
from openhands.tools.file_editor import FileEditorTool
from openhands.tools.terminal import TerminalTool

llm = LLM(
    model=os.getenv("LLM_MODEL", "anthropic/claude-sonnet-4-5-20250929"),
    api_key=os.getenv("LLM_API_KEY"),
    base_url=os.getenv("LLM_BASE_URL", None),
)
agent = Agent(llm=llm, tools=[Tool(name=TerminalTool.name), Tool(name=FileEditorTool.name)])
conversation = Conversation(agent=agent, workspace=os.getcwd())
conversation.send_message("List all Python files and count lines of code.")
conversation.run()

The LLM class accepts model, api_key, and base_url constructor arguments and can be configured via the LLM_MODEL, LLM_API_KEY, and LLM_BASE_URL environment variables to reach any LiteLLM-compatible provider. It also accepts openhands/-prefixed model names (e.g. openhands/claude-sonnet-4-5-20250929) when using OpenHands Cloud as the LLM provider.

$ export LLM_MODEL="openhands/claude-sonnet-4-5-20250929"
export LLM_API_KEY="your-openhands-api-key"
uv run python examples/01_standalone_sdk/01_hello_world.py

import os
from openhands.sdk import LLM, Agent, Conversation, Tool
from openhands.tools.file_editor import FileEditorTool
from openhands.tools.terminal import TerminalTool

llm = LLM(
    model=os.getenv("LLM_MODEL", "anthropic/claude-sonnet-4-5-20250929"),
    api_key=os.getenv("LLM_API_KEY"),
    base_url=os.getenv("LLM_BASE_URL", None),
)
agent = Agent(llm=llm, tools=[Tool(name=TerminalTool.name), Tool(name=FileEditorTool.name)])
conversation = Conversation(agent=agent, workspace=os.getcwd())
conversation.send_message("List all Python files and count lines of code.")
conversation.run()

LLM.subscription_login(vendor, model) authenticates with a ChatGPT Plus/Pro subscription so agents can access Codex models (e.g. gpt-5.2-codex) without consuming API credits.

from openhands.sdk import LLM
llm = LLM.subscription_login(vendor="openai", model="gpt-5.2-codex")

OpenHands ships as the openhands-sdk and openhands-tools PyPI packages, which must be installed together at matching versions (e.g. pip install "openhands-sdk==1.22.1" "openhands-tools==1.22.1"); optional openhands-workspace and openhands-agent-server packages add sandboxed workspaces in Docker or on a remote server. The SDK can also be built from source via git clone https://github.com/OpenHands/software-agent-sdk.git and make build.

Agent Canvas connects to execution, automation, and sandbox services with documented architecture, and can run Claude Code, Codex, or Gemini CLI inside it via the Agent Client Protocol (ACP). It can connect to a remote Agent Server backend on another machine or container, run with Docker for a sandboxed backend with a mounted project workspace, and manage LLM profiles to configure and reuse models across conversations.

claude remote-control --continue resumes the most recent Remote Control session without specifying a session ID, useful for reconnecting after a network drop. ListAgents now marks disconnected Remote Control sessions as offline and labels cloud sessions as cloud, giving clearer session inventory.

$ claude remote-control --continue

Introduces workflow fan-out staggering that lets same-prefix sibling agents share a cached prompt prefix; it can be disabled by setting CLAUDE_CODE_WORKFLOW_PREFIX_STAGGER_MS=0, e.g. for latency-sensitive pipelines where agents should not wait before starting.

$ CLAUDE_CODE_WORKFLOW_PREFIX_STAGGER_MS=0 claude

Self-hosted runner Windows startup now requires an explicit --base-dir; there is no default checkout directory on Windows, so launches that relied on a default will fail without it.

/commit-push-pr no longer auto-approves git/gh commands with dangerous flags such as --force, --amend, --no-verify; workflows relying on silent approval of these flags will now be prompted for confirmation.

Adds plugin marketplace command sources: a local command (e.g. an IDE) prints the plugin directory, which is re-resolved each session without a restart. mode: "link" uses the directory in place rather than copying it.

The HTTP MCP transport (skillspector mcp with http transport) now rejects local filesystem scan targets and local YARA-rule directories, enforcing a security boundary between remote callers and scanner-host paths. The stdio transport retains full local-scan capability; HTTP clients must switch to a remote repository or URL, or use stdio for local scans.

$ skillspector mcp --transport stdio

Adds a HIGH-severity SC8 finding when a skill ships Python bytecode or __pycache__ content, expanding supply-chain coverage.

Adds SettingDefinitionBase#disabled to the plugin API, letting plugin authors programmatically disable individual settings controls.

Images in Live Preview can now be selected via keyboard, no longer auto-expand to show filenames, and pressing Backspace/Delete removes the embedded image when selected.