The Cyber Toolchain — issue 047, August 16, 2026

New GET /v1/incidents/secrets/{incident_id}/locations endpoint lists every file on a source's default branch still holding a secret for a given incident, with exact line positions, filterable by source, number_of_mentions_min, number_of_mentions_max, code_fix_request_status (processing, created, error), and search (filepath substring), sortable by last_updated_at, number_of_mentions, or source. A companion GET /v1/sources/{source_id}/locations endpoint lists all files in a single source with any secret on the default branch, carrying the same per-match line positions, filters, and sort options.

$ curl -s -H 'Authorization: Token <api_key>' 'https://api.gitguardian.com/v1/incidents/secrets/42/locations?monitored=true&number_of_mentions_min=2&ordering=-number_of_mentions'

$ curl -s -H 'Authorization: Token <api_key>' 'https://api.gitguardian.com/v1/sources/5523/locations?search=settings.py&code_fix_request_status=created'

The ggshield honeytoken plant command (via the existing agent binary) plants a decoy AWS credential on developer endpoints; any use of the decoy triggers an alert naming the compromised machine, its IP address, user agent, and timestamp, with no new agent required. A new Honeytoken protection card in the Endpoints dashboard reports fleet coverage percentage and lists unprotected endpoints.

$ ggshield honeytoken plant

New beta GET /v1/scans/{scan_id}/live-progress endpoint returns the live status (queued, in_progress, completed) and real-time counters (elements_scanned, elements_total, elements_skipped, elements_failed, data_downloaded_bytes, data_scanned_bytes) for a running historical scan, including an audit_url field once the scan completes.

$ curl -s -H 'Authorization: Token <api_key>' 'https://api.gitguardian.com/v1/scans/<scan_id>/live-progress'

A new API endpoint lets teams push their own validity status (valid or invalid) for secrets on unreachable internal systems, authenticated with a personal access tokenpersonal access tokenA personal access token, a credential used by scripts and integrations instead of a user's password. Its permissions and expiry depend on the service issuing it. scoped to secrets:write; the status propagates to all incidents for that secret, appears as Custom validity with a datestamp, and is recorded in the incident activity log and audit log.

The API now supports creating Personal Access Tokens (PATsPATsA personal access token, a credential used by scripts and integrations instead of a user's password. Its permissions and expiry depend on the service issuing it.) and Service Account Tokens (SATs) programmatically for automated token rotation workflows; tokens created via the API must have a finite TTLTTLTime to live, a limit on how long data remains valid or how far a packet can travel. Caches use a time limit; IP packets use a hop limit.. Scans triggered via a Personal Access Token now also emit an audit log event.

Extends ignoreip with file:... (or file://path/file-with-ip-set) syntax to load ignored IPs, subnets, DNS/FQDN, or raw strings from an external file. The file is read lazily and reloaded dynamically when it changes, with comments after # or ; ignored; implemented via the new FileIPAddrSet class in ipdns.py.

# In jail.local or jail.conf
[DEFAULT]
ignoreip = 127.0.0.1/8 ::1 file:///etc/fail2ban/trusted-ips.txt

Adds filter.d/exim.conf mode more (several rules split out from mode normal) and extends mode aggressive to catch ACLACLAn access control list, a set of rules allowing or denying operations for particular users, services or network addresses. It decides who can use a resource.-dropped failures such as ACL: Country is banned. Also extends filter.d/sshd.conf modes ddos and aggressive with regex for timeout-before-authentication connections.

# In jail.local
[exim]
enabled = true
filter  = exim[mode=aggressive]

allow_file_and_shm_inputs is added explicitly to the normal, controller, and proxy worker configs (local.d/worker-normal.inc, local.d/worker-controller.inc, local.d/worker-proxy.inc) to gate TCPTCPA network protocol that delivers a reliable, ordered stream of bytes between two endpoints. It detects lost data and retransmits it. access to File, Path, and Shm message inputs; unix-socket connections always retain the capability, and the option must not be exposed to untrusted networks. It defaults to true in 4.1.5 with a startup warning on every TCP listener, but will default to false in the next major release, so TCP clients supplying file paths or shared-memory inputs must migrate to a unix socket before then. The proxy also strips client-supplied Shm* headers and File, Path, Shm query arguments, so scanning local files through the proxy no longer works and such clients must connect directly to a scanner listener.

allow_file_and_shm_inputs = true;

# local.d/worker-normal.inc
allow_file_and_shm_inputs = false;

The fuzzy_nested option in local.d/elastic.conf switches the fuzzy results field from a flat fuzzy_hashes list to the Elasticsearch nested type, with per-match objects from task:get_fuzzy_results() containing rule, symbol, found, queried hashes, type, probability/prob, flag, and exactness/exact; the schema also adds an explicit mapping for message size. Saved searches, dashboards, and ingest pipelines referencing the old flat field must be updated — indices created after the upgrade use the new mapping shape while pre-upgrade indices retain the old one.

fuzzy_nested = true;

Rules sharing extensions with their storage now attach a sender-facts extension to fuzzy check commands sent over encrypted connections, carrying SPFSPFAn email authentication standard that lets a domain owner publish, via DNS, which mail servers are authorized to send on its behalf, so receiving servers can reject spoofed senders., DKIMDKIMAn email authentication standard that lets a sending mail server cryptographically sign outgoing messages so receiving servers can verify the signature was not forged or tampered with in transit., and DMARC results, PTR state (including generic/dynamic hostname verdict from hfilter), bucketed recipient count, and TLS usage; FUZZY_CALLBACK gains ordering dependencies so authentication results are available when commands are built. A new no_share = true per-rule flag in local.d/fuzzy_check.conf disables sender-facts sharing on specific fuzzy rules. Fuzzy storage nodes older than 4.1.5 cannot skip this new extension type and will reject the entire fuzzy command with no reply to scanners — all fuzzy storage nodes must be upgraded to 4.1.5 before scanners, or no_share = true set on affected rules.

# local.d/fuzzy_check.conf, inside the affected rule
no_share = true;

max_auth_failures and auth_failure_window in local.d/worker-controller.inc rate-limit failed controller password attempts per source (default: 10 failures in 60s triggers HTTP 429; set max_auth_failures = 0 to disable). max_connections and max_connections_per_source limits are also added to the controller and proxy workers.

# local.d/worker-controller.inc
max_auth_failures = 20;
auth_failure_window = 120s;

The gpt module's condition callback in local.d/gpt.conf now returns only a boolean verdict plus an optional reason string, instead of both a decision and request content; request content generation moves internally into gpt_check. Deployments whose condition callback also returned content will have that content silently ignored and must update it to return the boolean decision alone.

encrypted_only = true in local.d/worker-fuzzy.inc no longer exempts peers with local or link-local source addresses — all clients, including those on localhost, must now use encryption (encryption_key in their fuzzy_check rule) or encrypted_only must be removed.

Per-key forbidden_ids ACLs in fuzzy storage no longer additively merge with the worker-level forbidden_ids option — keys that previously relied on the global list to suppress additional flags must have those flag IDs added explicitly to their own forbidden_ids list.

url:get_urls_filtered() called without an include list (used by rspamadm mime urls and the URI_COUNT_ODD rule) no longer drops flag-less URLs such as those extracted from HTML parts, correcting the parity check for multipart/alternative messages. As a result, HFILTER_HELO_IP_A and URI_COUNT_ODD now fire in cases where they previously did not — local score overrides and composites involving these symbols may produce different scores after upgrade.

Adds font-aware PDF simple font text decoding to libmime, supporting the three Annex D base encodings and per-font /Differences overrides, with a builder producing a single UTF-8 buffer per page. Also adds PDF /ToUnicode CMap parsing for composite (Type0) fonts, supporting codespace ranges, bfchar, bfrange, and UTF-16BE destinations including surrogate pairs; the CMap takes precedence over /Encoding.

Adds a Layers tab to the Resource Details drawer for container image resources, showing the image layer stack grouped by base image or Application Layers, with the Dockerfile instruction and per-severity vulnerability counts per layer. Also adds a Layer filter to the CVE list in Vulnerability Findings so findings can be sliced directly by image layer.

KSPMKSPMKubernetes Security Posture Management: a category of tooling that continuously audits Kubernetes clusters against configuration best practices and compliance policies to surface misconfigurations before attackers exploit them. Analyzer v1.49.3 adds detection of Container Network Interface (CNI) plugins in use on each Kubernetes node, enabling automatic evaluation of CIS KubernetesCIS KubernetesHardening benchmarks published by the Center for Internet Security for Kubernetes clusters, giving cyber tools a concrete, versioned checklist of configuration checks to run and report against. Benchmark v2.0.0 recommendation 5.3.1 without manual attestation.

The run API now accepts background: true, returning a JSON handle so long-running sessions don't block the caller. New runs/{runId} and runs/{runId}/logs endpoints let callers poll status and stream logs for these background jobs.

$ # Start a background run
curl -X POST https://execd/run \
  -H 'Content-Type: application/json' \
  -d '{"background": true, "command": "python train.py"}'

# Poll status using the returned runId
curl https://execd/runs/<runId>

# Stream logs
curl https://execd/runs/<runId>/logs

All five SDKs gain attach(sessionId), which rebuilds a fully functional session handle from a bare session ID (e.g. after a worker restart). SessionState now echoes the session's original creation parameters.

session = client.attach(sessionId)

execd now independently probes setpriv/userns support via uid_mode availability checks, returning 503 NOT_SUPPORTED for unavailable isolation modes instead of failing all isolation globally when one mode is unsupported.

Sessions created with share_net: false now bind-pin the workload NetNS/UserNS and return a retryable 503 on failure; previously a bind-pin failure did not block session creation, but now it fails closed at startup.

A new ?mode=viewer query parameter on PTY sessions lets multiple read-only clients watch a live PTY alongside the exclusive holder, supporting replay, live output, and backpressure isolation between viewers.

Adds query params for filtering/pagination to a wide set of list endpoints: GET /actors (email, name, type), GET /clients (firezone_id, name), GET /groups (directory_id, entity_type, name), GET /policies (group_id, resource_id), GET /resources (address, ip_stack, name, site_id, type), GET /sites (name), GET /sites/{site_id}/gateways (ipv4, ipv6, name), and the identity provideridentity providerA service that authenticates users and supplies identity information to other applications. It is the sign-in authority behind many single sign-on systems./directory list endpoints GET /email_otp_auth_providers, GET /entra_auth_providers, GET /entra_directories, GET /google_auth_providers, GET /google_directories, GET /oidc_auth_providers, GET /okta_auth_providers, and GET /okta_directories, each of which also gained limit and page_cursor params with a corresponding response schema change for cursor-based pagination.

Adds POST /sites/{site_id}/gateways to create a gateway within a site, plus PATCH /sites/{site_id}/gateways/{id} and PUT /sites/{site_id}/gateways/{id} to update an existing gateway; the corresponding GET/DELETE /sites/{site_id}/gateways/{id} response schemas also changed to match.

Adds GET /intune_devices, GET /intune_devices/{id}, and GET /intune_integration endpoints, providing read access to Microsoft Intune-managed device records and integration status via the Firezone REST API.

Adds GET /resources/{resource_id}/pool_members, PATCH /resources/{resource_id}/pool_members, and PUT /resources/{resource_id}/pool_members to read and update the pool members associated with a resource.

A new in-browser nudge, configured under Settings > In-browser nudges, intercepts sensitive data (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., secrets, financial info) detected in AI-tool conversations, shows the employee exactly what was found, and offers one-click redaction. An optional enforcement mode blocks sending until the sensitive data is redacted.

The new Dashboards > Passkey usage dashboard, populated by the browser extension, surfaces passkey adoption rate, top apps using passkeyspasskeysPhishing-resistant sign-in credentials built on FIDO2/WebAuthn: the device holds a private key and proves possession instead of the user typing a password. They sync across a user's devices., a storage provider breakdown toggleable between passkey count and user count, device-loss survivability, and per-login verification status. Passkey usage is also now shown per individual employee on their user details page.

The new Dashboards > Risk dashboard (open beta) shows residual risk scores, a trend chart filterable to business-critical apps, a business-criticality-vs-residual-risk matrix with drill-down, and a ranked highest-risk-apps table built automatically from Nudge's per-app risk scoring.

codex doctor now checks free disk space for both CODEX_HOME and the current worktree, showing a warning below a 5 GiB threshold and failing below 1 GiB; on Windows targets it also checks for a Dev Drive.

Threads can now start without blocking while environment attachment configuration is still pending, while ensuring turns do not use an attachment prematurely; adds a Failed configuration state and an environment_failed callback for per-thread failure and recovery.

Propagates W3C traceparent and tracestate trace context through exec-server relay frames, including fragmented encrypted Noise requests.

Chunked SSM re-derive ships enabled by default, restoring long-context cache reuse from 0% to 99.9% above 12.5k tokens on hybrid SSM model families. This was validated with a 100k-context multiturn test where a 97.6k-token conversation retrieved planted facts with 99.9% cache reuse and no memory faults.

A hybrid prefix cache delivers up to 20x faster follow-up turns on long documents: a 43.7k-token document follow-up drops from 107s to 5s, with output byte-identical at temperature 0.

Privacy Mode changes session visibility defaults so only sessions with policy violations are automatically surfaced to security teams, while compliant sessions stay private unless an authorized user records an explicit justification for access. Certiv logs who accessed a private compliant session and the recorded reason, creating an auditable trail for governance, privacy teams, and works councils. It integrates with the existing Policy Engine and Scout endpoint observer, so policy verdict determines session visibility without altering real-time Block, Pause, and Steer interventions. Privacy Mode is now available in Preview to existing Certiv customers, enabled on request through the Certiv team.

⤢

Adds opt-in auto-discovery of a top-level .skillspector-baseline.yaml file, keeping explicitly supplied baselines authoritative when both are present.