Ivanti Appliance Exploitation and Persistence

Critical severity TLP:CLEAR Part 1 of 2 Research by Rapid7 Published 18 Nov 2026

Hypothesis

An intruder has exploited a zero-day vulnerability chain on an Ivanti Connect Secure appliance to gain root access, evidenced by X-DSCGI-PATH header anomalies followed by the execution of rare python payloads and selective log tampering.

Brief

Why Now

Rapid7 recently published details on Additional Ivanti Connect Secure zero-days exploited in the wild. The report highlights a new vulnerability chain where attackers bypass existing integrity checks to achieve root access. Because these actors actively tamper with local logs and bypass standard detection tools, a behavioral hunt across multiple telemetry surfaces is necessary to confirm if an appliance is compromised.

How the Hunt Flows

The hunt begins by scoping the estate for appliances that remain vulnerable to the reported CVEs. It checks vulnerability finding data to identify hosts that have not yet applied the 25.1.1.3 or 22.8R2.6 updates. This ensures the analyst focuses subsequent high-volume queries on the most likely targets.

Next, the hunt searches for early access indicators. It runs two parallel queries: one for HTTP requests containing the X-DSCGI-PATH header in the log lines or URL queries, and another for network connections to known adversary command-and-control infrastructure. If an analyst finds these signals, it suggests the exploitation attempt succeeded.

The third phase investigates appliance-internal activity. The hunt looks for rare python processes executed from temporary or data directories, specifically focusing on filenames like ml.py or rq5.py. Simultaneously, it searches for the deletion of system audit and debug logs, which the adversary uses to hide their presence and activities after gaining root access.

In the final phase, an analyst correlates the initial web exploit leads with the observed root execution and log tampering. This unified view allows for a high-confidence verdict on whether the appliance has been fully compromised by UTA0576, facilitating immediate containment actions like network isolation.

What the Hunt Cannot See

This hunt has two primary blind spots. First, it relies on the appliance exporting process-level telemetry to an EDR or a central logging server. If the appliance is a closed system that does not provide internal process logs, the execution of python payloads will remain invisible. Second, many web server configurations do not log custom headers by default. If the HTTP activity logs omit the X-DSCGI-PATH header, the primary indicator of initial access will be missed.

In this series

Steps

  1. Query, scoping

    Identify Vulnerable Ivanti Appliances

    Scope the estate for appliances that remain vulnerable to the exploitation chain.

    sql, reads hb_vulnerability_finding
    SELECT resource_uid, cve_uid, severity, statusFROM hb_vulnerability_findingWHERE instr(',' || '{{target_cves}}' || ',', ',' || cve_uid || ',') > 0  AND status != 'suppressed'ORDER BY severity_id DESC

    What a hit looks like. A list of resource UIDs representing Ivanti appliances still exposed to the CVEs. Silence implies no known vulnerable assets in the current inventory.

  2. Query, detection candidate

    Exploitation Header Signals

    Identify web requests containing the X-DSCGI-PATH header associated with the Ivanti exploit chain.

    sql, reads hb_http_activity
    SELECT device_hostname, url_path, url_query, log_line, src_endpoint_ip, timeFROM hb_http_activityWHERE (url_path LIKE '%welcome.cgi%' OR url_query LIKE '%welcome.cgi%')  AND (LOWER(log_line) LIKE '%x-dscgi-path%' OR LOWER(url_query) LIKE '%x-dscgi-path%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. HTTP requests targeting the vulnerable endpoint with specific path injection headers. Silence proves no such header was observed in captured logs.

  3. Query, enrichment

    Adversary Infrastructure Connections

    Identify connections from the appliance to known attacker IP addresses.

    sql, reads hb_network_connection
    SELECT device_hostname, dst_endpoint_ip, dst_endpoint_port, direction, timeFROM hb_network_connectionWHERE instr(',' || '{{attacker_ips}}' || ',', ',' || dst_endpoint_ip || ',') > 0  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. Any connection from an appliance to the listed C2 infrastructure. Silence means no traffic to these indicators was recorded.

  4. Agent triage

    Early Stage Exploitation Assessment

    Evaluate whether the HTTP header signals and network connections indicate successful initial access.

  5. Query, baseline

    Rare Root Payloads in Temp Paths

    Identify rare process executions from temporary paths or using known malicious filenames.

    sql, reads hb_process_activity
    SELECT LOWER(process_path) AS path, process_cmd_line, COUNT(DISTINCT device_hostname) AS hosts, MIN(time) AS first_seenFROM hb_process_activityWHERE (instr(',' || '{{malicious_payload_patterns}}' || ',', ',' || LOWER(process_name) || ',') > 0 OR process_path LIKE '%/tmp/%' OR process_path LIKE '%/data/%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')GROUP BY 1, 2HAVING hosts <= 5ORDER BY hosts ASC

    What a hit looks like. Processes running as root or from temporary directories seen on very few appliances. Silence proves no such rare execution was captured in the telemetry.

  6. Query, triage

    Selective Log Deletion

    Detect deliberate removal of critical Ivanti debug and audit logs used to hide activity.

    sql, reads hb_file_activity
    SELECT device_hostname, file_path, activity_name, process_name, timeFROM hb_file_activityWHERE activity_id = 4  AND (file_path LIKE '%/home/runtime/dlogs/debuglog%' OR file_path LIKE '%/var/log/audit/audit.log%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. Deletion events for system-level audit logs. Silence means no deletions of these specific files were logged.

  7. Agent triage

    Full Attack Chain Analysis

    Correlate early exploit leads with root execution and log tampering to confirm compromise.

  8. Decision

    Route on Final Verdict

    Route to immediate containment for malicious findings or analyst review for suspicious signals.

  9. Response action

    Isolate Malicious Appliance

    Isolate the compromised appliance to prevent lateral movement and further credential theft.

  10. Analyst task

    Manual Forensic Review

    An analyst reviews the correlated rows to confirm exploitation and identify any stolen secrets.

  11. Analyst task

    Remediation and Close-out

    Finalize remediation steps and close the hunt.

Coverage

4 of 5 stages covered

Stage and ATT&CK techniqueCoveredHow, or why not
Web Application Exploitation
T1190
Yes exploit-header-signals
Root-Level Payload Execution
T1059.006 T1068
Yes rare-root-payloads
Defense Evasion and Log Tampering
T1070.002
Yes selective-log-deletion
Command and Control Traffic
T1071.001
Yes network-c2-connections
Active Directory Secret Discovery
T1087.002 T1212
Out of scope Belongs to another part of the 'Additional Ivanti Connect Secure zero-days exploited in the wild (CVE-2026-111357, CVE-2026-118527)' series.

Blind spots

  • Needs EDR or native Linux audit telemetry from the ICS appliance. If the appliance does not export process-level telemetry, root payload execution will be invisible to this hunt. It would answer: Can we see internal process execution on the closed appliance?
  • Needs hb_http_activity with custom header logging. Many standard access log configurations omit custom headers; if the header is not captured, the primary exploit indicator is missed. It would answer: Does the web server log the X-DSCGI-PATH header?

Parameters and data

Parameters

ParameterTypeDefaultWhat it is
attacker_ipslist[ip]108.165.147.64Known attacker IP addresses from Volexity research.
lookback_daysnumber14Days of history to examine.
malicious_payload_patternslist[string]ml.py, rq5.py, mk.py, rr.py, h.pyKnown malicious filenames dropped by the threat actor.
scope_hostslist[host]noneOptional list of hostnames to scope the hunt; leave empty for fleet-wide.
target_cveslist[string]CVE-2026-104935, CVE-2026-111357, CVE-2026-118527Target CVE identifiers for the Ivanti exploit chain.

Data sources

SourceCategoryTelemetry
Endpoint telemetry (hb_ surfaces)endpointendpoint
Network telemetrynetworknetwork
Web server / proxy logssiemnetwork

Source

An open hunt.md file. It runs anywhere that reads the format.

hunt.md
---
analysis: This hunt correlates edge exploit signals (HTTP headers) with follow-on
  appliance-internal activity (rare root processes and log deletion). A single rule
  cannot link these disparate telemetry sources into a unified attack chain narrative,
  especially when binaries are rare and paths rotate.
blind_spots:
- id: appliance-telemetry-gap
  question: Can we see internal process execution on the closed appliance?
  requires: EDR or native Linux audit telemetry from the ICS appliance
  risk: If the appliance does not export process-level telemetry, root payload execution
    will be invisible to this hunt.
  stage: execution-root-payloads
- id: header-logging-missing
  question: Does the web server log the X-DSCGI-PATH header?
  requires: hb_http_activity with custom header logging
  risk: Many standard access log configurations omit custom headers; if the header
    is not captured, the primary exploit indicator is missed.
  stage: initial-access-web-exploitation
coverage:
- stage: initial-access-web-exploitation
  status: covered
  steps:
  - exploit-header-signals
- stage: execution-root-payloads
  status: covered
  steps:
  - rare-root-payloads
- stage: defense-evasion-selective-deletion
  status: covered
  steps:
  - selective-log-deletion
- stage: c2-external-comms
  status: covered
  steps:
  - network-c2-connections
- reason: Belongs to another part of the 'Additional Ivanti Connect Secure zero-days
    exploited in the wild (CVE-2026-111357, CVE-2026-118527)' series.
  stage: discovery-ad-secrets
  status: out_of_scope
guardrails:
  claims: no_unsupported
  evidence: citation_required
  missing_data: not_benign
  telemetry: untrusted
hunt:
  applicability: campaign-specific
  handoff: keep-as-periodic-hunt
  justification: Ivanti appliances are critical perimeter assets frequently targeted
    by zero-day chains. With reports of UTA0576 bypassing standard integrity checks,
    a multi-surface behavioral hunt is required to ensure the integrity of the network
    edge.
  methodology: model-assisted
  trigger: intel-report
hypothesis: An intruder has exploited a zero-day vulnerability chain on an Ivanti
  Connect Secure appliance to gain root access, evidenced by X-DSCGI-PATH header anomalies
  followed by the execution of rare python payloads and selective log tampering.
labels:
- hunt
- attack.t1190
- attack.t1059.006
- attack.t1068
- attack.t1070.002
- attack.t1071.001
- command and control
- defense evasion
- discovery
- execution
- initial access
name: Ivanti Appliance Exploitation and Persistence
parameters:
  attacker_ips:
    default:
    - 108.165.147.64
    description: Known attacker IP addresses from Volexity research.
    from:
      kind: article
      observed: '2026-11-02'
      ref: Rapid7
    type: list[ip]
  lookback_days:
    default: '14'
    description: Days of history to examine.
    from:
      kind: manual
      observed: '2026-11-02'
      ref: hunt-standard
    type: number
  malicious_payload_patterns:
    default:
    - ml.py
    - rq5.py
    - mk.py
    - rr.py
    - h.py
    description: Known malicious filenames dropped by the threat actor.
    from:
      kind: article
      observed: '2026-11-02'
      ref: Rapid7
    type: list[string]
  scope_hosts:
    default: []
    description: Optional list of hostnames to scope the hunt; leave empty for fleet-wide.
    from:
      kind: manual
      observed: '2026-11-02'
      ref: analyst-defined
    type: list[host]
  target_cves:
    default:
    - CVE-2026-104935
    - CVE-2026-111357
    - CVE-2026-118527
    description: Target CVE identifiers for the Ivanti exploit chain.
    from:
      kind: article
      observed: '2026-11-02'
      ref: Rapid7
    type: list[string]
provenance:
  authors:
  - name: Huntbase hunt generation
    org: huntbase.io
  generated:
    by: huntbase-hunt-generation
    from: https://www.rapid7.com/blog/post/etr-additional-ivanti-connect-secure-zero-days-exploited-in-the-wild-cve-2026-111357-cve-2026-118527
    gates:
    - dry-run
    - lint
    - critic
    model: hb_google/gemini-3-flash-preview
rationale: Target all Ivanti Connect Secure (ICS) appliances. Prioritize appliances
  running versions older than 25.1.1.3 or 22.8R2.6 identified in the vulnerability
  scan data.
references:
- name: "Rapid7 \u2014 Additional Ivanti Connect Secure zero-days exploited in the\
    \ wild"
  url: https://www.rapid7.com/blog/post/etr-additional-ivanti-connect-secure-zero-days-exploited-in-the-wild-cve-2026-111357-cve-2026-118527
related:
- hunt: discovery-ad-secrets
  reason: Hunting for the theft of LAPS or BitLocker secrets from Active Directory
    requires examining Domain Controller logs, which is a separate identity-focused
    hunt.
  relation: out-of-scope-alternative
scenario:
  stages:
  - name: Web Application Exploitation
    observables:
    - /dana-na/auth/url_default/welcome.cgi
    - 'X-DSCGI-PATH: /home/webserver/htdocs/dana/fb/smb/wu.cgi'
    - 'X-DSCGI-PATH: /home/runtime/tmp/'
    - CVE-2026-104935
    - CVE-2026-118527
    slug: initial-access-web-exploitation
    tactic: initial-access
    techniques:
    - T1190
  - name: Root-Level Payload Execution
    observables:
    - ml.py
    - rq5.py
    - mk.py
    - rr.py
    - pu.txt
    - sp_out.txt
    - spray_hits.json
    - dsstatmond
    - ics-root-run-inband-*.pl
    - CVE-2026-111357
    slug: execution-root-payloads
    tactic: execution
    techniques:
    - T1059.006
    - T1068
  - name: Defense Evasion and Log Tampering
    observables:
    - /home/runtime/dlogs/debuglog
    - /var/log/audit/audit.log
    - /data/.icscap/
    - h.py
    slug: defense-evasion-selective-deletion
    tactic: defense-evasion
    techniques:
    - T1070.002
  - name: Active Directory Secret Discovery
    observables:
    - Active Directory queries for Windows LAPS passwords
    - Active Directory queries for BitLocker recovery keys
    slug: discovery-ad-secrets
    tactic: discovery
    techniques:
    - T1087.002
    - T1212
  - name: Command and Control Traffic
    observables:
    - 108.165.147.64
    slug: c2-external-comms
    tactic: c2
    techniques:
    - T1071.001
  summary: The threat actor UTA0576 exploited a chain of three zero-day vulnerabilities
    in Ivanti Connect Secure (CVE-2026-104935, CVE-2026-111357, and CVE-2026-118527)
    to gain unauthenticated root access. After establishing a foothold, they deployed
    Python payloads to steal credentials and queried Active Directory for high-value
    targets including LAPS passwords and BitLocker recovery keys.
series:
  index: 1
  slug: additional-ivanti-connect-secure-zero-days-exploited-in-the-wild-cve-2026-111357-cve-2026-118527
  title: Additional Ivanti Connect Secure zero-days exploited in the wild (CVE-2026-111357,
    CVE-2026-118527)
  total: 2
severity: critical
targets:
  analyst:
    name: Tier-2 analyst
    role: analyst
  endpoint:
    category: endpoint
    name: Endpoint telemetry (hb_ surfaces)
    telemetry:
    - endpoint
  hunter:
    agent: true
    name: Hunt agent
  network:
    category: network
    name: Network telemetry
    telemetry:
    - network
  web:
    category: siem
    name: Web server / proxy logs
    telemetry:
    - network
tlp: clear
type: investigation
---


# Ivanti Appliance Exploitation and Persistence

This hunt identifies the exploitation of Ivanti Connect Secure (ICS) appliances by tracking the UTA0576 attack chain. It follows a phased approach, first scoping the estate for vulnerable versions and then searching for initial access indicators in HTTP headers and network connections to known adversary infrastructure.

If early signals are present, the hunt pivots to examine root-level post-exploitation activity on the appliance itself. This includes searching for rare python scripts executed from temporary directories and identifying attempts to delete system audit and debug logs. This multi-surface correlation is designed to overcome the limitations of the vendor's integrity checker, which the actor is known to bypass.

## identify-vulnerable-appliances
<!-- Identify Vulnerable Ivanti Appliances -->
Scope the estate for appliances that remain vulnerable to the exploitation chain.

```sqlite target=endpoint role=scoping params=(target_cves=target_cves)
~~~yaml
expected: A list of resource UIDs representing Ivanti appliances still exposed to
  the CVEs. Silence implies no known vulnerable assets in the current inventory.
reads:
- resource_uid
- cve_uid
- severity
- status
silence: not_evidence_of_absence
source: hb_vulnerability_finding
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT resource_uid, cve_uid, severity, status FROM hb_vulnerability_finding WHERE instr(',' || '{{target_cves}}' || ',', ',' || cve_uid || ',') > 0 AND status != 'suppressed' ORDER BY severity_id DESC
```

## early-stage-fan-out
<!-- Concurrent Early Stage Monitoring -->
parallel:
- → exploit-header-signals
- → network-c2-connections
join: → early-stage-triage

## exploit-header-signals
<!-- Exploitation Header Signals -->
Identify web requests containing the X-DSCGI-PATH header associated with the Ivanti exploit chain.

```sqlite target=web role=detection-candidate params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: HTTP requests targeting the vulnerable endpoint with specific path injection
  headers. Silence proves no such header was observed in captured logs.
reads:
- device_hostname
- url_path
- url_query
- log_line
- src_endpoint_ip
- time
silence: not_evidence_of_absence
source: hb_http_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, url_path, url_query, log_line, src_endpoint_ip, time FROM hb_http_activity WHERE (url_path LIKE '%welcome.cgi%' OR url_query LIKE '%welcome.cgi%') AND (LOWER(log_line) LIKE '%x-dscgi-path%' OR LOWER(url_query) LIKE '%x-dscgi-path%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## network-c2-connections
<!-- Adversary Infrastructure Connections -->
Identify connections from the appliance to known attacker IP addresses.

```sqlite target=network role=enrichment params=(lookback_days=lookback_days, attacker_ips=attacker_ips, scope_hosts=scope_hosts)
~~~yaml
expected: Any connection from an appliance to the listed C2 infrastructure. Silence
  means no traffic to these indicators was recorded.
reads:
- device_hostname
- dst_endpoint_ip
- dst_endpoint_port
- direction
- time
silence: not_evidence_of_absence
source: hb_network_connection
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, dst_endpoint_ip, dst_endpoint_port, direction, time FROM hb_network_connection WHERE instr(',' || '{{attacker_ips}}' || ',', ',' || dst_endpoint_ip || ',') > 0 AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## early-stage-triage
<!-- Early Stage Exploitation Assessment -->
```agent target=hunter
cite: required
context:
- exploit-header-signals
- network-c2-connections
max_iterations: 3
objective: Determine if the Ivanti appliances show evidence of exploitation through
  header injection or known C2 traffic.
success_criteria: A verdict for each affected host citing specific rows from HTTP
  and network sources.
tools:
- endpoint
- network
- web
```

## follow-on-fan-out
<!-- Concurrent Post-Exploitation Monitoring -->
parallel:
- → rare-root-payloads
- → selective-log-deletion
join: → full-attack-chain-analysis

## rare-root-payloads
<!-- Rare Root Payloads in Temp Paths -->
Identify rare process executions from temporary paths or using known malicious filenames.

```sqlite target=endpoint role=baseline params=(lookback_days=lookback_days, malicious_payload_patterns=malicious_payload_patterns, scope_hosts=scope_hosts)
~~~yaml
baseline:
  compare: first_seen
  window: '{{lookback_days}}d'
expected: Processes running as root or from temporary directories seen on very few
  appliances. Silence proves no such rare execution was captured in the telemetry.
prevalence:
  by: device_hostname
  key:
  - process_path
  - process_cmd_line
  rare_below: 5
reads:
- device_hostname
- process_path
- process_name
- process_cmd_line
- time
silence: not_evidence_of_absence
source: hb_process_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT LOWER(process_path) AS path, process_cmd_line, COUNT(DISTINCT device_hostname) AS hosts, MIN(time) AS first_seen FROM hb_process_activity WHERE (instr(',' || '{{malicious_payload_patterns}}' || ',', ',' || LOWER(process_name) || ',') > 0 OR process_path LIKE '%/tmp/%' OR process_path LIKE '%/data/%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY 1, 2 HAVING hosts <= 5 ORDER BY hosts ASC
```

## selective-log-deletion
<!-- Selective Log Deletion -->
Detect deliberate removal of critical Ivanti debug and audit logs used to hide activity.

```sqlite target=endpoint role=triage params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: Deletion events for system-level audit logs. Silence means no deletions
  of these specific files were logged.
reads:
- device_hostname
- file_path
- activity_id
- activity_name
- process_name
- time
silence: not_evidence_of_absence
source: hb_file_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, file_path, activity_name, process_name, time FROM hb_file_activity WHERE activity_id = 4 AND (file_path LIKE '%/home/runtime/dlogs/debuglog%' OR file_path LIKE '%/var/log/audit/audit.log%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## full-attack-chain-analysis
<!-- Full Attack Chain Analysis -->
```agent target=hunter
cite: required
context:
- early-stage-triage
- rare-root-payloads
- selective-log-deletion
max_iterations: 4
objective: Assess if the appliance has moved from initial web exploitation (Agent
  1) to root payload execution and log evasion.
success_criteria: A final verdict citing the correlated chain of evidence from initial
  access to post-exploitation.
tools:
- endpoint
- network
- web
```

## route-on-verdict
<!-- Route on Final Verdict -->
if~: "The final attack chain analysis verdict is malicious for at least one Ivanti appliance." (confidence: high, judge=hunter)
then: → isolate-appliance
indeterminate: → manual-forensic-review
unavailable: → manual-forensic-review (blind_spot: appliance-telemetry-gap)
else: → manual-forensic-review

## isolate-appliance
<!-- Isolate Malicious Appliance -->
```action target=endpoint
~~~yaml
approval: required
~~~
Isolate the identified Ivanti appliance and prepare to capture memory and disk images for deep forensics. Avoid rebooting until memory has been preserved.
```
→ manual-forensic-review

## manual-forensic-review
<!-- Manual Forensic Review -->
```manual target=analyst
1. Review cited rows in the full-attack-chain-analysis step. 2. Verify if Active Directory queries for LAPS passwords or BitLocker keys were initiated from this host. 3. Confirm whether the appliance's local logs show evidence of deletion.
```
→ remediation-and-close-out

## remediation-and-close-out
<!-- Remediation and Close-out -->
```manual target=analyst
1. Apply updates 25.1.1.3 or 22.8R2.6 to all appliances. 2. Rotate credentials for any service accounts active on the appliance during the breach window. 3. Document the final state of the fleet.
```
→ end

Run this hunt against your own data.

In Huntbase, Scout runs every step across the tools you already connect, checks back through your history and returns a verdict with the evidence. The hunt.md file is open, so you can also take it anywhere that reads the format.

Drafted by machine (huntbase-hunt-generation, hb_google/gemini-3-flash-preview), checked by dry-run, lint, critic, then reviewed by a person.

Content pack

Sign up to open

Pack contents open in Huntbase.