The Cyber Toolchain — issue 035, August 4, 2026

New --project-memory-preload-limit flag sets the number of memories preloaded at session start (range: 1–200), replacing a hardcoded default of 20.

$ dn ... --project-memory-preload-limit 100

New dn airt run-classifier CLI command runs classic-ML attacks — evasion, extraction, membership inference, and model inversion — directly against a predict endpoint without SDK code.

Adds four traditional-ML attack families (evasion, extraction, membership inference, model inversion) to AIRT, covering 20+ named attacks — including equation-solving, Jacobian, copycat, knockoff, PWWS, BAE, TextBugger, threshold, and label-only — across tabular, image, and text modalities, with per-finding metrics, adversarial trajectory traces, and MITRE ATLASMITRE ATLASA MITRE-maintained knowledge base of adversarial tactics and techniques targeting AI and machine learning systems, modelled on ATT&CK, giving cyber tools a structured taxonomy for AI-specific threat modelling./SAIF/NIST compliance mappings. Model inversion attacks reconstruct class-representative inputs via black-box queries to surface privacy risks, completing the fourth traditional-ML attack category.

Adds /report-bug command (or r in console) to the TUITUIA text-based user interface that runs inside a terminal, using characters and ANSI escape codes to render menus, panes, and widgets without a graphical desktop, letting cyber tools ship rich interactivity with no GUI dependency., assembling a privacy-reviewed diagnostic bundle and submitting it without losing logs after a crash.

$ /report-bug

New dn evaluation create --agent <name> flag specifies which agent runs in an evaluation; eval output reports which agent actually ran.

$ dn evaluation create --agent <name>

Adds dn capability sync and dn task sync CLI commands to pull hub content from SaaSSaaSSoftware delivered over the internet and run on the provider's infrastructure, so the buyer pays a subscription instead of installing or operating anything locally. (or another on-prem) into a local directory and push it to a self-hosted instance, enabling air-gapped on-prem loading.

Agent Output now functions as a severity-first triage workspace — searches, sorts, and filters an engagement's findings; each finding can be exported to CSV, deep-linked, or copied into a report.

Adds ASTASTAn abstract syntax tree, a structured representation of parsed code. Tools inspect or transform this tree to understand expressions and statements rather than just matching text.-based JS/TS static analysis via ast-grep in the web-security runtime, removing the need for semgrep, CodeQL, or a jxscout license.

Adds the ATLAS multi-agent attack framework for running adaptive campaigns against deployed multi-agent environments across three attack surfacesattack surfacesThe exposed systems, interfaces and access points an attacker could try to use. It includes internet-facing services as well as identities, software and other reachable resources., with executed tool calls captured per-trial and surfaced in findings across the UI, API, and SDK.

Extraction findings include KL divergence and accuracy retention; membership inference includes attack accuracy, precision, and recall; model inversion includes reconstruction quality.

Adds the llm-wiki capability (compiles source material into linked wiki pages) and analysis-ledger capability (maintains a resumable code-analysis frontier with evidence-bearing claims).

Session monitoring redesigned: adds a Last Msg relative-activity column, highlights sessions active in the last 15 minutes, and adds forward/backward transcript pagination with tailing mode.

New web-security initial-recon skill provides structured reconnaissance for domain, wildcard, ASNASNAutonomous System Number — a globally unique identifier assigned by regional internet registries to a network operator, letting routing protocols distinguish which organization controls a block of IP addresses., IP-list, and CIDRCIDRA notation for a block of IP addresses, such as 10.0.0.0/24. The number after the slash says how many leading address bits identify the network. entry points using the ProjectDiscovery toolchain.

Five self-contained AIRT demo notebooks cover model evasion, extraction, membership inference, jailbreaks, and multi-agent attacks — runnable via dn login with no provider keys required.

Sessions group-by control (None / Workflow / Capability / Agent / Source) is now available to all orgs, with workflow runs collapsed by default.

Agent Output items from failed report_item POSTs are automatically recovered from span backups when a session is frozen.

New Memories page under Agents lets operators view, inspect, and manage project memory entries, including version history and diffs.

Stat cards for Needs review and Critical + High on the Agent Outputs page are now clickable filter toggles; the items grid gains a sortable Updated column.

Rapid Response now surfaces personalized asset-level exploitability status, remediation guidance, and verification feedback to distill vulnerability noise. The Rapid Response Details page shows an Asset Breakdown, CVE or H3 IDs, and a Relevant Assets table with filtering, Action Items and Fix Actions guidance, and a Take Action button offering options to run a Rapid Response Test, add a note, create a ticket, or run an external or internal test.

⤢ ⤢ ⤢ ⤢

Tripwires deploys silent decoy credentials — real credentials generated by Horizon3 that blend into the environment — to assets flagged as high-risk by NodeZero autonomous pentest results; tokens stay silent until triggered, then alert Horizon3's cloud infrastructure. Tripwires now also supports CSV export, credential rotation, and improved Active DirectoryActive DirectoryMicrosoft's directory service for Windows networks that centralises authentication and authorisation; cyber tools target or integrate with it because it controls identity and access across most enterprise environments. validation.

A new No Longer Found status was added to the Vulnerability Management Hub (VMH), with enhanced sorting and filtering by Host, Weakness ID, and Downstream Impact.

External Pentest was rebuilt, replacing Asset Groups with reusable Scopes, improving bulk asset management and uploads, and enhancing support for large external attack surfacesattack surfacesThe exposed systems, interfaces and access points an attacker could try to use. It includes internet-facing services as well as identities, software and other reachable resources..

Client deletion now returns scanned asset allocation to the parent within 30 days, down from a previous 60-day hold.

Credentials stored in AWS Secrets Manager, Azure Key Vault, or HashiCorp Vault can be mapped to integrations and virtual keys via secret_mappings, with Portkey fetching values at runtime instead of storing them directly.

⤢

{
  "virtual_key": "my-openai-key",
  "secret_mappings": {
    "api_key": "arn:aws:secretsmanager:us-east-1:123456789012:secret:prod/openai-api-key"
  }
}

Adds GCP_WIF_AUDIENCE and GCP_WIF_SERVICE_ACCOUNT_EMAIL environment variables to enable GCP Workload Identity Federation (gcs_assume style) for GCS log delivery when the gateway runs in AWS, eliminating static cross-cloud credentials.

$ GCP_WIF_AUDIENCE=//iam.googleapis.com/projects/123456/locations/global/workloadIdentityPools/my-pool/providers/my-provider \
GCP_WIF_SERVICE_ACCOUNT_EMAIL=portkey-log-writer@my-project.iam.gserviceaccount.com \
docker run portkey/gateway ...

Applies Zscaler Detection Policies to LLM inputs and outputs using beforeRequestHook and afterRequestHook hooks, requiring a policyId and supporting a configurable timeout (default 10000 ms).

Adds Vertex AI enterprise web search grounding via enterpriseWebSearch / enterprise_web_search with separate cost attribution, Portkey metadata mapping to Vertex resource labels, and AWS-to-Vertex access via GCP_WIF_AUDIENCE and GCP_WIF_SERVICE_ACCOUNT_EMAIL.

Adds stream_options support for DeepSeek deepseek-chat, and maps reasoning_effort to thinking mode for deepseek-reasoner, returning reasoning_content in streams.

Adds rpw (requests-per-week) budget and usage windows so rate limits can align with weekly planning cycles, alongside endpoint-scoped limits that let chat completions, embeddingsembeddingsLists 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., and admin routes carry independent caps.

⤢

Scans LLM inputs and outputs for 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 sensitive data leakage via hooks with a configurable timeout (default 5000 ms).

Adds DeepInfra provider support with tool callingtool callingA model's ability to answer with a structured request to run a named function or API, which the application executes and returns. It is how agents take actions. via tools, tool_choice, and parallel_tool_calls, plus completions and embeddings endpoints.

Adds Azure AI Foundry rerank support for Cohere rerank models (e.g. cohere.Cohere-rerank-v4.0-pro), with the gateway stripping the cohere. prefix for the provider.

OpenTelemetryOpenTelemetryAn open standard and set of tools for collecting traces, metrics and logs from applications. It lets services export observability data without tying instrumentation to one monitoring vendor. GenAI semantic spans now follow semconv 1.40.0 for inference and embeddings, with OTELOTELAn open standard and set of tools for collecting traces, metrics and logs from applications. It lets services export observability data without tying instrumentation to one monitoring vendor. exporter support for guardrail flows and custom resource attributes.

skill-scanner-pre-commit --install installs a Git pre-commit hook that maps each changed path to its nearest parent SKILL.md, scans each affected skill once, and blocks commits at the configured severity; --scan-all scans every configured skill instead of only affected ones. Incremental CI scans are supported via pre-commit run skill-scanner --from-ref "$BASE_SHA" --to-ref "$HEAD_SHA", scanning only files changed between two revisions.

$ skill-scanner-pre-commit --install

$ pre-commit run skill-scanner --from-ref "$BASE_SHA" --to-ref "$HEAD_SHA"

--adjudicate flag / adjudicator.enabled and adjudicator.min_fp_confidence policy keys enable a demote-only LLM review that can reduce deterministic HIGH/CRITICAL literal-regex false positives to INFO without ever promoting a finding; disabled by default and records decisions for audit. SKILL_SCANNER_ADJUDICATOR_LLM_MODEL overrides the model used specifically for this path (e.g. vertex_ai/gemini-2.5-pro).

adjudicator:
  enabled: true
  min_fp_confidence: 4

$ skill-scanner scan ./my-skill --adjudicate

$ skill-scanner scan ./skill --use-llm --enable-meta --adjudicate

build_analyzers in skill_scanner.core.analyzer_factory assembles optional analyzers (use_behavioral, use_llm, use_osv, llm_consensus_runs) programmatically, matching CLI and API parity; the demote-only adjudicator pass can be enabled via policy.adjudicator.enabled = True before constructing SkillScanner.

from skill_scanner.core.analyzer_factory import build_analyzers
analyzers = build_analyzers(
    policy,
    use_behavioral=True,
    use_llm=True,
    use_osv=True,
    llm_consensus_runs=3,
)
scanner = SkillScanner(analyzers=analyzers, policy=policy)

policy.adjudicator.enabled = True
scanner = SkillScanner(analyzers=analyzers, policy=policy)

--rule-packs selects named packs; the atr (Cisco AI Threat Research) pack expanded from 314 to 712 rules (pack version 3.5.6), and the promptguard pack ships 26 rules covering 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., provider secrets, and LLM/Markdown exfiltration. --rule-packs list enumerates all bundled packs.

$ skill-scanner scan ./skill --rule-packs atr promptguard

$ skill-scanner scan /path/to/skill --use-behavioral --use-osv --use-llm --enable-meta --rule-packs atr promptguard

Repeating --format (e.g. --format json --format sarif) produces multiple report formats in a single scan run, paired with per-format output path flags --output-json, --output-sarif, --output-markdown, --output-html, and --output-table, eliminating the need to scan twice for different report formats. --render-markdown / --no-render-markdown override terminal Markdown rendering.

$ skill-scanner scan ./my-skill --format json --format sarif \
  --output-json results.json --output-sarif results.sarif

$ skill-scanner scan /path/to/skill --format json --format sarif --output-json results.json --output-sarif results.sarif

New scan-repo subcommand clones a public GitHub repository (accepts owner/repo shorthand or a full HTTPSHTTPSHTTP carried over an encrypted TLS connection. It protects traffic in transit and lets a browser verify the server's certificate. URL) into a temp directory, recursively discovers and scans every skill in it, then removes the clone; --no-recursive disables recursive discovery (recursive is on by default).

$ skill-scanner scan-repo https://github.com/owner/repo --use-osv --check-overlap

$ skill-scanner scan-repo owner/repo

--use-osv flag / use_osv config field queries the OSV.dev advisory database for vulnerabilities in exactly pinned Python dependencies, requiring network access but no API key or additional package.

use_osv: true
use_trigger: true
llm_consensus_runs: 3

$ skill-scanner scan-repo https://github.com/owner/repo --use-osv --check-overlap

$ skill-scanner scan ./skill --use-osv

$ skill-scanner scan /path/to/skill --use-osv

An openai-compatible LLM provider shortcut plus SKILL_SCANNER_LLM_PROVIDER (set to openai), SKILL_SCANNER_LLM_BASE_URL to point at any OpenAI-compatible base URL, SKILL_SCANNER_LLM_USER to inject a raw Chat Completions user field, and SKILL_SCANNER_LLM_TEMPERATURE (or none to omit sampling temperature) configure routing to OpenAI-compatible endpoints.

llm_analysis.trusted_reference_domains config key lists organization-controlled domains whose transitive-trust and supply-chain findings are automatically demoted to LOW severity; described elsewhere as allowlisting domains that should not be flagged as suspicious during semantic analysis.

llm_analysis:
  trusted_reference_domains:
    - git.example.com
    - packages.example.com

The reusable workflow's CLI allowlist now includes --rule-packs, --use-aidefense, --use-trigger, and --enable-meta, plus LLM limits and repository-local policy/taxonomy paths; unknown flags are rejected before the scan runs. --use-osv and --adjudicate are not accepted by the reusable workflow and require the self-hosted pattern instead.

llm_consensus_runs integer config field (default 1) controls the number of LLM runs used for majority consensus adjudication, and is also exposed programmatically through build_analyzers.

use_osv: true
use_trigger: true
llm_consensus_runs: 3

from skill_scanner.core.analyzer_factory import build_analyzers
analyzers = build_analyzers(
    policy,
    use_behavioral=True,
    use_llm=True,
    use_osv=True,
    llm_consensus_runs=3,
)
scanner = SkillScanner(analyzers=analyzers, policy=policy)

Scan results include an aggregate llm_usage field with input_tokens, output_tokens, and total_tokens when the configured provider reports usage, across LLM, meta, and adjudication calls, enabling cost tracking and auditing; also exposed as a dict[str, int] field in the Python SDK.

Secure Registry now serves NuGet (Beta) at https://registry.stepsecurity.io/dotnet/index.json (V3 protocol only, resources under /dotnet/v3/), routing all NuGet metadata requests and package downloads through Secure Registry for policy evaluation. Policy controls available at launch are Cooldown Period, Compromised Packages, and Custom Block List (TyposquattingTyposquattingUsing a name close to a trusted domain or software package to catch spelling mistakes. Attackers use it to lure people or installers to a malicious lookalike. Protection is Coming Soon). Supports three integration paths — JFrog Artifactory, Sonatype Nexus Repository, and Direct (with a Package manager selector switching setup instructions between .NET CLI, NuGet CLI, and Visual Studio) — and NuGet now appears in the Registry selector on both the Policy and Setup Guide tabs, configured independently per ecosystem.

Adds package_config device policy support: the agent converges a StepSecurity-owned block in ~/.npmrc, bounded by BEGIN/END markers, to route npm, pnpm, Yarn classic, and bun through the tenant's Secure Registry using a per-device ::dev:-suffixed authentication token; absent policy is a no-op with snapshot rollback, and surrounding configuration is left untouched.

Introduces a verify-only enforcement channel that reads the OS-managed VS Code policy (Windows registry, macOS managed preferences, /etc/vscode/policy.json) or effective ~/.npmrc and reports observed values and drift without writing. MDM verification performs field-by-field drift diffs, reporting one of eight states (MDM verified, MDM drift, Not deployed, Verification failed, and others) with desired vs. observed values per field. Every profile now declares an explicit enforcement type — Verify only (MDM-enforced) or Self healing (agent-enforced) — and switching an existing profile's type migrates all assigned devices, which report Pending until each agent checks in.

Secure Registry does not yet proxy the VulnerabilityInfo resource for NuGet, causing audit-enabled restores to emit NU1905; the documented workaround is to pass -p:NuGetAudit=false or exclude the Secure Registry source from audit sources, e.g. dotnet restore -p:NuGetAudit=false.

$ dotnet restore -p:NuGetAudit=false

Writes extensions.gallery.serviceUrl into VS Code user-scope settings.json atomically alongside extensions.allowed when a policy sets a gallery URL; removal is ownership-gated so developer-configured URLs are never deleted. IDE extension policies now support an optional gallery URL that points assigned devices at a self-hosted extension gallery instead of the public Visual Studio Marketplace, and this gallery URL is itself verified for drift.

NuGet is now selectable from the Package ecosystem toggle in OSS Package Search; packages are indexed from *.csproj, *.vbproj, *.fsproj, packages.config, packages.lock.json, Directory.Packages.props, Directory.Build.props, and *.deps.json files.

GitHub Workflow Integration v2.1.1 updates the default runner image to scm-packaging-5.0.0, adding Alpine v3.24.0, Ubuntu v26.04, .NET v10, PipEnv v2026.6.2, libicu v78, Node.js v22.22.1, and Python v3.14.4.

The new --enterprise-db-token TOKEN flag (also settable via the WPSCAN_ENTERPRISE_DB_TOKEN environment variable or the enterprise_db_token config-file key) lets WPScan scan against a local vulnerability database dump instead of the online API, with no per-request limits. When set, WPScan downloads plugins.json.gz, themes.json.gz, and wordpresses.json.gz from enterprise-data.wpscan.org, authenticated via the X-DB-JSON-AUTH header, and reads vulnerability data locally; API status is then reported as plan enterprise with Unlimited requests remaining. --enterprise-db-token and --api-token are mutually exclusive (including via their environment variables) — supplying both raises an error — and the token is redacted from logged command lines.

$ wpscan --url https://target.tld --enterprise-db-token <your-enterprise-token> --enumerate vp,vt

cli_options:
  enterprise_db_token: "<your-enterprise-token>"
  url: "https://target.tld"

New extract-base64 subcommand scans .evtx or JSON logs for Base64-encoded content and saves results to a CSV file via -o <FILE>; it scans PowerShell event IDs 4100/4102 (Microsoft-Windows-PowerShell/Operational and PowerShellCore/Operational) across ContextInfo and Payload fields, and classic Windows PowerShell EIDs 403/600 via the EventData.Data blob.

$ hayabusa.exe extract-base64 -d ../hayabusa-sample-evtx -o base64-extracted.csv

New dfir-timeline command replaces csv-timeline and json-timeline; output format is selected via -t, --output-type: csv (default), json, or jsonl (case-insensitive, e.g. -t JSONL). The -t short flag is reassigned from --threads to --output-type (use --threads <N> for thread count), the old -L, --JSONL-output flag is removed in favor of -t jsonl, CSV-only flags (-M, --multiline, -S, --tab-separator, -R, --remove-duplicate-data) now error if combined with a non-CSV -t value, and the scan wizard is enabled by default.

$ hayabusa.exe dfir-timeline -d .\hayabusa-sample-evtx -t JSONL -o results.jsonl

-R, --remove-duplicate-data replaces duplicate field data with "DUP" in dfir-timeline output (requires --sort); -X, --remove-duplicate-detections strips duplicate detections from dfir-timeline (requires --sort); -X, --remove-duplicate-records deduplicates event records in eid-metrics and logon-summary (default: disabled).

$ hayabusa.exe dfir-timeline -E -s -X -d .\hayabusa-sample-evtx -o results.csv

logon-summary now counts RDP session events: Security EIDs 4778/4779 (session reconnect/disconnect with client workstation name and IP) and Terminal Services events RemoteConnectionManager/Operational 1149 and LocalSessionManager/Operational 25, covering RDP activity even when 4624 events have been overwritten.

All long option names are renamed to lowercase: --geo-ip, --HTML-report→--html-report, --UTC→--utc, --ISO-8601→--iso-8601, --RFC-2822/--RFC-3339→--rfc-2822/--rfc-3339, --JSON-input/--JSON-output→--json-input/--json-output, --EID-filter→--eid-filter; short flags are unchanged and scripts using the old uppercase forms will break.

-G, --geo-ip <MAXMIND-DB-DIR> enriches IP addresses in dfir-timeline output with GeoIP data (ASNASNAutonomous System Number — a globally unique identifier assigned by regional internet registries to a network operator, letting routing protocols distinguish which organization controls a block of IP addresses., city, country).

$ hayabusa.exe dfir-timeline -d .\hayabusa-sample-evtx -G .\maxmind-db -o results.csv

-E, --eid-filter restricts scanning to common Event IDs listed in ./rules/config/target_event_IDs.txt for faster performance; added to dfir-timeline and pivot-keywords-list.

$ hayabusa.exe dfir-timeline -E -s -X -d .\hayabusa-sample-evtx -o results.csv

Adds --european-time, -O, --iso-8601 (always UTC), --rfc-2822, --rfc-3339, -U, --utc, --us-military-time, and --us-time timestamp format flags across dfir-timeline, eid-metrics, extract-base64, log-metrics, logon-summary, and search.

$ hayabusa.exe log-metrics -d ../logs --include-channel Security,System -U -o log-metrics.csv

--exclude-channel <CHANNEL...>, --exclude-filename <FILE...>, --include-channel <CHANNEL...>, and --include-filename <FILE...> flags added to log-metrics for fine-grained channel and file filtering.

$ hayabusa.exe log-metrics -d ../logs --include-channel Security,System -U -o log-metrics.csv

-s, --sort sorts dfir-timeline results before saving (note: significantly increases memory usage); -T, --visualize-timeline outputs an event-frequency timeline in the terminal (requires unicode terminal support and --sort).

-M, --multiline separates event field information by newline characters (CSV only), added to dfir-timeline, log-metrics, and search; -S, --tab-separator separates fields by tabs (CSV only) in dfir-timeline; combining either with a non-CSV -t value produces an error.

-H, --html-report <FILE> saves a Results Summary HTML report alongside the dfir-timeline output.

New Linux aarch64 (ARM64) musl release binary (hayabusa-<ver>-lin-aarch64-musl) built with a proper aarch64 musl cross toolchain.

-J, --json-input scans JSON-formatted logs (.json or .jsonl) instead of .evtx in dfir-timeline.

--threads <NUMBER> controls thread count (default: optimal for performance), added to dfir-timeline, eid-metrics, extract-base64, log-metrics, logon-summary, pivot-keywords-list, and search.

-V, --validate-checksums enables checksum validation, added to dfir-timeline, computer-metrics, eid-metrics, extract-base64, log-metrics, logon-summary, pivot-keywords-list, and search.

-J, --json-output and -L, --jsonl-output flags save search results in JSON or JSONL format.

Adds --dump-table metadata dump mode to ilspycmd for inspecting raw metadata tables, e.g. ilspycmd --dump-table MyAssembly.dll.

$ ilspycmd --dump-table MyAssembly.dll

Adds --member flag to ilspycmd for single-member decompilation, e.g. ilspycmd --member "MyNamespace.MyClass.MyMethod" MyAssembly.dll, useful for quick diffs in CI.

$ ilspycmd --member "MyNamespace.MyClass.MyMethod" MyAssembly.dll

Run phishing simulation campaigns prioritized by attack patterns Sublime is already detecting in your environment. Supports multiple templates per campaign with staggered delivery across recipients, and tracks click rates, report rates, and repeat clickers in a cross-campaign dashboard with recipient-level detail per campaign.

⤢

A new Terraform template provisions the 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. role and policies as native Terraform resources and registers the AWS integration automatically, requiring no CloudFormationCloudFormationAn AWS service that lets you describe and provision cloud infrastructure as declarative YAML or JSON templates, so a cyber tool can reproducibly spin up or tear down its entire AWS environment with a single API call. stack or manual console step. It offers a choice of running workload scanning inside your own account or through a Plerion-managed service account.

Workload scanning now builds a software bill of materialssoftware bill of materialsA 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. (SBOM) for every container image across Kubernetes Deployments, DaemonSets, StatefulSets, Pods, Jobs, and CronJobs, surfaced on each asset's SBOM tab.

Onboards an entire Google Cloud organization from a single generated command using keyless access via Workload Identity Federation, with no service account keys required. An equivalent Terraform template is also provided for organizations preferring infrastructure-as-code onboarding.

Secret scanningSecret scanningChecking files, code or other data for exposed credentials such as passwords, API keys and tokens. Findings help teams remove the secret and revoke or rotate it. detects API keys, tokens, and passwords inside container images and reports them with severity on the asset's Sensitive data tab, so practitioners know what to rotate first.

The new Data Exposure Audit uses 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. agents to trace sensitive data from sources (request bodies, uploads, DB reads, config, application logic) to sinks (logs, third-party APIs, LLM prompts, caches, queues, storage) entirely from code, with no production data store connection required. It detects credential and token leakage in URLs, query strings, redirects, and clone URLs; API over-exposure where endpoints return more fields than the caller is entitled to or lack tenant ownership checks; unsafe logging and exports of unredacted 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. and secrets in logs, traces, caches, and dumps; over-broad third-party and AI data sharing to vendors, analytics, LLM prompts, and vector storesvector storesA database built to store embeddings and find the nearest ones to a query quickly. It is the retrieval half of most RAG systems.; weak storage patterns such as regulated data stored unencrypted or protected only with base64, MD5, or SHA-1SHA-1A cryptographic hash function standardized by NIST that produces a 160-bit digest; tools use it for file integrity checks, though it is now considered cryptographically broken and unsuitable for collision-resistant applications.; and incomplete deletion flows that leave downstream records, caches, shadow copies, staging data, or snapshots intact. Each Data Exposure Audit run consumes credits from the Aikido Wallet.

⤢

Prowler CLI adds a new huaweicloud provider authenticated via HUAWEICLOUD_ACCESS_KEY_ID and HUAWEICLOUD_SECRET_ACCESS_KEY environment variables, running prowler huaweicloud across 25 checks covering CTS, 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., ELB, EVS, 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., KMSKMSA managed service that creates, stores, rotates and controls access to encryption keys, usually backed by HSMs. Applications ask it to encrypt or decrypt rather than holding keys themselves., OBS, RDS, 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., and WAFWAFA reverse proxy that inspects HTTP/HTTPS traffic and blocks requests matching known attack patterns such as SQL injection or XSS, sitting between clients and a web application to filter malicious input before it reaches the server. services, plus the CIS Huawei Cloud Foundations Benchmark 1.0 compliance framework.

$ HUAWEICLOUD_ACCESS_KEY_ID="your-access-key-id" HUAWEICLOUD_SECRET_ACCESS_KEY="your-secret-access-key" prowler huaweicloud

$ export HUAWEICLOUD_ACCESS_KEY_ID="your-access-key-id"
export HUAWEICLOUD_SECRET_ACCESS_KEY="your-secret-access-key"
prowler huaweicloud

Prowler MCP Cloud Server at https://mcp.prowler.com/mcp adds 32 tools in the prowler_cloud_* namespace covering Alerts (rule creation/management, dry-run, fired-alert history), Findings Triage (status and notes), Scan Scheduling (daily, interval, weekly, or monthly per-provider or bulk), and Scan Configurations (reusable check/compliance selections attached to providers).

Adds 11 new checks: five ec2_confidential_workload_host_* checks (IMDSv2 enforcement, public IP exposure, unrestricted ingress, exposed vsock proxy ports, stopped hosts) and six kms_key_enclave_* checks (attestation enforcement, bypass, deployment binding, debug-mode, PCR mismatches, unknown enclave images), mapped across 23 compliance frameworks including NIST 800-53NIST 800-53A catalog of security and privacy controls for U.S. federal information systems, published by NIST. Cyber tools cite it to map their findings or features to a recognized compliance framework. Rev 5, 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. v4.0, ISOISOA widely adopted international standards body that publishes technical specifications; cyber tools cite ISO standards to signal compliance with recognized benchmarks for quality, security, or interoperability. 27001:2022, SOC 2SOC 2An auditing framework from the American Institute of CPAs that certifies a service organization's controls around security, availability, and data handling, giving buyers assurance that a vendor meets baseline trust criteria., HIPAAHIPAAA US federal law enacted in 1996 that sets privacy and security requirements for protected health information, giving cyber tools a compliance baseline to enforce or audit against., and MITRE ATT&CK.

Lighthouse AI in Prowler Cloud gains a context-aware chat panel that automatically attaches the current app page and any open finding or resource metadata to each message — shown as a chip in the composer — so queries like explain this resolve against what is on screen, and Lighthouse can now invoke all Prowler MCP ServerMCP ServerA process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. tool families (scan configurations, scheduling, finding triage, alert rules, integrations, users, roles), gated by RBAC.

⤢ ⤢ ⤢ ⤢ ⤢ ⤢

Adds four new Attack Paths detection queries: STS-002 (cross-account role trust), STS-003 (wildcard role trust), IAM-022 (user permissions-boundary removal), and SSO-001 (IAM Identity Center permission-set escalation).

Prowler MCP Server (Cloud and self-hosted Local) gains three tool families: Integrations (full lifecycle for Amazon S3, AWS Security Hub, and Jira — create, update, re-check, delete, and dispatch findings as Jira work items), Users (list tenant users and identify the authenticated user), and Roles (browse RBAC roles, inspect capabilities, and set a user's role).

⤢