db
database
PostgreSQL 16 database for Windmill, configured with a healthcheck.
labs/windmill/docker-compose.yml:14-23Chocapikk/Windfall
PoC files
67 filesFile viewing is interactive and short-lived. Downloads are password-protected ZIP archives using password eip.
Analysis
Technical assessment
The repository 'Windfall' contains a full exploitation framework for CVE-2026-29059 (path traversal) and CVE-2026-23696 (SQL injection) in Windmill/Nextcloud Flow. It includes Python scripts (windfall_afr.py, windfall_sqli.py, windfall_nc_pivot.py) and Metasploit modules that actively exploit these vulnerabilities to leak credentials, forge JWTs, and achieve remote code execution.
Backdoor review
No backdoor observed in reviewed code
The supplied evidence consists of documentation and Metasploit module source code for a publicly disclosed path traversal and SQL injection exploit (CVE-2026-29059, CVE-2026-23696) targeting Windmill/Nextcloud Flow. All reviewed files describe or implement the expected exploit behavior: reading arbitrary files, leaking credentials, forging JWTs, and achieving remote code execution on the target. No concealed backdoor, unrelated payload, or operator-directed harm beyond the stated exploit was observed.
ClassificationExploit
Model confidence95%
AuthenticationNot required
Languages
Target software
Attack types
Evidence & reasoningClassification basis · observed behavior · safety review
Classification basis and observed behavior
Classification basis
The artifact includes Metasploit modules (windmill_path_traversal_rce.rb, windmill_sqli_rce.rb) that actively exploit vulnerabilities to achieve remote code execution. The README describes these as 'exploits' and provides usage examples for obtaining shells. The code is intended to exercise the vulnerability, not merely detect it.
README.md:128-130metasploit/modules/exploits/linux/http/windmill_path_traversal_rce.rb:17-21metasploit/modules/exploits/linux/http/windmill_sqli_rce.rb:18-22Requirements
- Target running vulnerable Windmill (v1.309.0 - v1.603.2) or Nextcloud Flow (v1.0.0 - v1.2.2)
README.md:23-24 - Network access to the target's Windmill API endpoint
README.md:195
Observed behavior
- Sends crafted HTTP requests with path traversal sequences (e.g., '../') to the vulnerable get_log_file endpoint to read arbitrary files like /etc/passwd.
README.md:294 - Leaks credentials by reading /proc/1/environ for SUPERADMIN_SECRET, windmill_users_config.json for tokens, or PostgreSQL data files for jwt_secret.
README.md:140-144 - Uses leaked credentials or forged JWT to authenticate to the Windmill API and execute arbitrary commands via the job execution feature, achieving remote code execution.
metasploit/modules/exploits/linux/http/windmill_path_traversal_rce.rb:115-117 - The SQL injection module authenticates as a low-privileged user, exploits a SQLi in the addowner endpoint to extract jwt_secret, forges an admin JWT, and escalates privileges to superadmin for RCE.
metasploit/modules/exploits/linux/http/windmill_sqli_rce.rb:147-165
Behaviors behind the backdoor verdict
Observables
- Exploit Behavior
- Path traversal to read /etc/passwd, /proc/1/environ, PostgreSQL data files, and windmill_users_config.jsonThis is the documented vulnerability exploitation path; no hidden behavior detected.
README.md:107-110metasploit/modules/exploits/linux/http/windmill_path_traversal_rce.rb:121-124 - Exploit Behavior
- Credential leak via SUPERADMIN_SECRET, windmill_users_config.json, and PostgreSQL jwt_secret extractionMatches the documented attack chain; no evidence of exfiltration to an attacker-controlled destination beyond the module's normal operation.
metasploit/modules/exploits/linux/http/windmill_path_traversal_rce.rb:134-145 - Exploit Behavior
- Remote code execution via Windmill job API using leaked or forged tokensThe module executes attacker-supplied commands on the target; this is the stated purpose of the exploit.
metasploit/modules/exploits/linux/http/windmill_path_traversal_rce.rb:115-117 - Exploit Behavior
- SQL injection via folder addowner endpoint to extract jwt_secret and escalate privilegesDocumented authenticated SQLi attack chain; no hidden queries or data exfiltration beyond the described escalation.
metasploit/modules/exploits/linux/http/windmill_sqli_rce.rb:147-165 - Opsec Technique
- Ghost mode cleanup (--clean) that deletes job traces from the databaseDescribed as an OPSEC feature to remove forensic evidence; this is part of the exploit's self-destruct mechanism, not a backdoor.
README.md:263-275
What the analysis did not establish
- Only 8 of 67 files in the repository were provided as text evidence. The actual Python exploit scripts (windfall_afr.py, windfall_sqli.py, windfall_nc_pivot.py) and other Metasploit module files are not included in the selected text.
- The evidence packet reports complete_artifact_coverage as false, indicating that not all files from the repository were analyzed.
- 59 files (344,110 bytes) were classified as unclassified or binary and were not inspected; their content is unknown.
- Only 8 of 67 total files were provided as readable text; the remaining files could contain additional code or configuration not reviewed.
Model interpretation
This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.
Docker lab environments
6GitHublabs/windmill/docker-compose.yml
Created Model review verdict: Suspicious behavior.Vuln labNo linked CVECompose · images
AnalysisSuspicious behavior
Environment assessment
A Docker Compose environment for Windmill (v1.603.2) consisting of a PostgreSQL database, a Windmill server, a Windmill worker with Docker socket access, and a setup service that creates a default operator user and workspace. The environment is designed to demonstrate path traversal, SQL injection, and host escape vulnerabilities.
labs/windmill/docker-compose.yml:1-91labs/windmill/README.md:1-43Lab assessment
Vulnerability labThe README explicitly states it is a 'Test environment for Windmill vulnerabilities' and provides commands to run path traversal, SQLi, and host escape exploits. The Docker Compose file pins a vulnerable version (1.603.2) and includes a comment about a Docker socket mount for a 'host escape demo (INSECURE - lab only!)'.
labs/windmill/README.md:1-3labs/windmill/docker-compose.yml:1-5labs/windmill/docker-compose.yml:62-63Lab shapeCompose · images
Services4
Compose manifests1
Dockerfiles0
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
windmill_server
web applicationvulnerable target
Windmill server v1.603.2, exposed on port 8000, with a SUPERADMIN_SECRET environment variable. It depends on the db service and mounts a volume for worker logs.
labs/windmill/docker-compose.yml:25-44windmill_worker
workerhost escape vector
Windmill worker v1.603.2, with the Docker socket mounted for a host escape demonstration. It depends on the db service.
labs/windmill/docker-compose.yml:46-63setup
initialization
A temporary service using curlimages/curl that creates an operator user, a demo workspace, and adds the operator to the workspace via the Windmill API, using the SUPERADMIN_SECRET for authentication.
labs/windmill/docker-compose.yml:66-89Requirements and sequence described by the evidence
Prerequisites
- Docker and Docker Compose installed on the host.
labs/windmill/README.md:7-8 - Python 3.10+ with dependencies (rich, requests, pyjwt, prompt_toolkit) installed, as specified in pyproject.toml.
pyproject.toml:6-12 - The exploit scripts (windfall_afr.py, windfall_sqli.py) must be present and runnable from the repository root.
labs/windmill/README.md:13-20 - The Docker socket (/var/run/docker.sock) must be accessible on the host for the host escape demonstration.
labs/windmill/docker-compose.yml:62-63
Evidence-described exercise path
- Start the lab environment with 'docker-compose up -d'.
labs/windmill/README.md:7-8 - Wait for services to become healthy.
labs/windmill/README.md:10-11 - Run the path traversal exploit: 'python3 windfall_afr.py http://localhost:8000 -c "id"'.
labs/windmill/README.md:13-14 - Run the SQL injection exploit: 'python3 windfall_sqli.py http://localhost:8000 -u operator@windmill.dev -p password123 -c "id"'.
labs/windmill/README.md:16-17 - Run the host escape exploit: 'python3 windfall_afr.py http://localhost:8000 --host'.
labs/windmill/README.md:19-20
Behaviors behind the stored safety assessment
Suspicious behavior
The windmill_worker service mounts the host's Docker socket (/var/run/docker.sock) for a documented host escape demonstration. While the README and Compose file disclose this as a lab-only insecure configuration, the mount enables the exploit to execute arbitrary commands on the Docker host, which constitutes concrete host impact. The behavior is disclosed, but the impact is severe and warrants a suspicious verdict.
labs/windmill/docker-compose.yml:62-63labs/windmill/README.md:19-20Indicators requiring review
- The windmill_worker service mounts the host Docker socket, enabling container escape and potential host compromise.
labs/windmill/docker-compose.yml:62-63
What the analysis did not establish
- The exploit scripts (windfall_afr.py, windfall_sqli.py) are not included in the evidence packet, so their exact behavior cannot be verified.
- The pyproject.toml references scripts that are not present in the packet, so the full exploit chain cannot be assessed.
- The CVE association is ambiguous; the repository claims CVE-2026-29059, but the unit has no CVE IDs, and no CVE details are provided in the evidence.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.
GitHublabs/postgres-post
Created UncertainNo linked CVEDockerfile
Analysis
Environment assessment
A PostgreSQL 16 container initialized with a SQL script that creates tables and inserts sample data, including simulated secrets, tokens, and credentials. The Dockerfile installs curl. The broader repository describes a Windmill/Nextcloud Flow vulnerability research project, but this unit only contains a standalone database setup.
labs/postgres-post/Dockerfile:1-6labs/postgres-post/init.sql:1-158Lab assessment
Uncertain lab purposeThe unit provides a PostgreSQL database with sample data that could serve as a target for SQL injection or credential theft exercises described in the repository README. However, the unit itself contains no exploit code, vulnerable application, or explicit vulnerability research instructions. It is a standalone database initialization that may support a lab but does not independently demonstrate a vulnerability research purpose.
labs/postgres-post/Dockerfile:1-6labs/postgres-post/init.sql:1-158README.md:1-490Lab shapeDockerfile
ServicesUnknown
Compose manifests0
Dockerfiles1
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
postgres:16 container
database server
A PostgreSQL 16 database server initialized with tables (users, secrets, api_tokens, metrics, files, events, sessions, tags, configurations) and sample data including simulated credentials, tokens, and secrets. The Dockerfile installs curl, but no curl usage is shown.
labs/postgres-post/Dockerfile:1-6labs/postgres-post/init.sql:1-158Requirements and sequence described by the evidence
Prerequisites
- Docker must be available to build the image from the Dockerfile and run the container.
labs/postgres-post/Dockerfile:1-6 - The init.sql script must be present in the build context to be copied into the container.
labs/postgres-post/Dockerfile:5
Behaviors behind the stored safety assessment
No harmful behavior observed
The unit only builds a PostgreSQL container with sample data. It does not execute any commands, expose ports, mount host resources, or perform any actions beyond database initialization. No harmful behavior is visible.
labs/postgres-post/Dockerfile:1-6labs/postgres-post/init.sql:1-158What the analysis did not establish
- The unit does not include a docker-compose.yml or any runtime configuration; port mappings, network settings, and volume mounts are unknown.
- The unit's relationship to the Windmill/Nextcloud Flow vulnerabilities described in the repository README is not defined within the unit itself.
- No exploit code or vulnerable application is present in this unit; the exercise path cannot be determined from the unit alone.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.
GitHublabs/postgres-post/docker-compose.yml
Created Not a vuln labNo linked CVECompose · images
Analysis
Environment assessment
A single-service Docker Compose environment running PostgreSQL 16 with a custom initialization script that creates tables and inserts sample data, including simulated secrets and credentials. The environment appears to be a standalone database lab, not directly related to the Windfall vulnerability research described in the repository README.
labs/postgres-post/docker-compose.yml:1-16labs/postgres-post/init.sql:1-158Lab assessment
Not identified as a vulnerability labThe Docker Compose file defines a basic PostgreSQL service with no vulnerable application, no exploit code, and no connection to the Windfall vulnerability research described in the repository README. The init.sql script populates the database with sample data, including fake secrets, but does not set up a vulnerable target or exercise any vulnerability. The lab is a simple database environment, not a vulnerability reproduction lab.
labs/postgres-post/docker-compose.yml:1-16labs/postgres-post/init.sql:1-158Lab shapeCompose · images
Services1
Compose manifests1
Dockerfiles0
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
postgres
database server
PostgreSQL 16 database service with environment variables for user, password, and database name. It mounts an initialization script and a named volume for data persistence. Port 15432 on the host is mapped to container port 5432.
labs/postgres-post/docker-compose.yml:2-12Requirements and sequence described by the evidence
Prerequisites
- Docker and Docker Compose must be installed to run the environment.
labs/postgres-post/docker-compose.yml:1 - The init.sql file must be present in the same directory as the Compose file.
labs/postgres-post/docker-compose.yml:9
Behaviors behind the stored safety assessment
No harmful behavior observed
The environment is a simple PostgreSQL database with no exploit code, no vulnerable application, and no behavior that targets the host, external systems, or persists beyond the container. The port mapping and volume mounts are standard for a database lab and do not indicate malicious intent.
labs/postgres-post/docker-compose.yml:1-16labs/postgres-post/init.sql:1-158What the analysis did not establish
- The repository README describes Windfall exploits for Windmill/Nextcloud Flow, but the lab unit contains only a PostgreSQL database with no Windmill or Nextcloud components. The relationship between this lab and the repository's vulnerability research is unclear.
- The unit_cve_ids array is empty, so no CVE assessments are provided. The README mentions CVE-2026-29059, CVE-2026-23696, and CVE-2026-22683, but these are not associated with this lab unit.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.
GitHublabs/postgres-multiversion/docker-compose.yml
Created UncertainNo linked CVECompose · images
Analysis
Environment assessment
A Docker Compose environment that starts six PostgreSQL database instances (versions 12 through 17) with a shared initialization script that creates tables and populates them with sample data including simulated secrets, tokens, and credentials.
labs/postgres-multiversion/docker-compose.yml:1-80labs/postgres-multiversion/init.sql:1-161Lab assessment
Uncertain lab purposeThe environment provides multiple PostgreSQL versions with a dataset that includes secrets, tokens, and password hashes, which could serve as a target for vulnerability research or exploit development. However, the Compose file and init script contain no exploit code, no vulnerable application, and no explicit vulnerability-research instructions. The README describes a Windmill/Nextcloud vulnerability lab, but the supplied unit (postgres-multiversion) does not include Windmill or Nextcloud components. The relationship between this database environment and the Windfall exploits is not established by the evidence.
labs/postgres-multiversion/docker-compose.yml:1-80labs/postgres-multiversion/init.sql:1-161README.md:1-490Lab shapeCompose · images
Services6
Compose manifests1
Dockerfiles0
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
pg12
PostgreSQL 12 database server
A PostgreSQL 12 container initialized with the shared init.sql script, exposing port 5432 on host port 15012.
labs/postgres-multiversion/docker-compose.yml:2-11pg13
PostgreSQL 13 database server
A PostgreSQL 13 container initialized with the shared init.sql script, exposing port 5432 on host port 15013.
labs/postgres-multiversion/docker-compose.yml:13-22pg14
PostgreSQL 14 database server
A PostgreSQL 14 container initialized with the shared init.sql script, exposing port 5432 on host port 15014.
labs/postgres-multiversion/docker-compose.yml:24-33pg15
PostgreSQL 15 database server
A PostgreSQL 15 container initialized with the shared init.sql script, exposing port 5432 on host port 15015.
labs/postgres-multiversion/docker-compose.yml:35-44pg16
PostgreSQL 16 database server
A PostgreSQL 16 container initialized with the shared init.sql script, exposing port 5432 on host port 15016.
labs/postgres-multiversion/docker-compose.yml:46-55pg17
PostgreSQL 17 database server
A PostgreSQL 17 container initialized with the shared init.sql script, exposing port 5432 on host port 15017.
labs/postgres-multiversion/docker-compose.yml:57-66Requirements and sequence described by the evidence
Prerequisites
- Docker and Docker Compose must be installed to start the services.
labs/postgres-multiversion/docker-compose.yml:1 - The init.sql file must be present in the same directory as the Compose file.
labs/postgres-multiversion/docker-compose.yml:8
Behaviors behind the stored safety assessment
No harmful behavior observed
The Compose file defines six standard PostgreSQL containers with no privileged mode, no host filesystem mounts beyond the init script, and no network exposure beyond mapped ports. The init script creates tables and inserts sample data; it does not execute shell commands, download payloads, or modify the host. No exploit code, backdoor, or destructive behavior is present in the supplied unit.
labs/postgres-multiversion/docker-compose.yml:1-80labs/postgres-multiversion/init.sql:1-161What the analysis did not establish
- The unit does not include the Windmill or Nextcloud components described in the repository README; the relationship between this database environment and the Windfall exploits is not established.
- The unit_cve_ids array is empty, so no CVE assessments are provided.
- No exercise path is described within the unit; the README describes exploits that target a different environment.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.
GitHublabs/nextcloud-flow/docker-compose.yml
Created Vuln labNo linked CVECompose · images
Analysis
Environment assessment
A Docker Compose environment for a Nextcloud AIO instance with an nginx reverse proxy, intended as a test lab for Windmill vulnerabilities via Nextcloud Flow.
labs/nextcloud-flow/README.md:1-3labs/nextcloud-flow/docker-compose.yml:1-44Lab assessment
Vulnerability labThe README explicitly states it is a 'Test environment for Windmill vulnerabilities via Nextcloud Flow' and provides exploit usage instructions for path traversal and SQLi.
labs/nextcloud-flow/README.md:1-3labs/nextcloud-flow/README.md:37-55Lab shapeCompose · images
Services2
Compose manifests1
Dockerfiles0
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
nextcloud-aio-mastercontainer
Nextcloud AIO master containerorchestrates Nextcloud services
Runs the Nextcloud All-in-One master container, which manages the Nextcloud instance. It mounts the Docker socket (read-only) and a named volume for configuration, and exposes port 8180 for the AIO web interface.
labs/nextcloud-flow/docker-compose.yml:7-19nginx-proxy
reverse proxySSL termination
An nginx container that terminates SSL on port 443 and proxies requests to the Nextcloud Apache service at nextcloud-aio-apache:11000. It uses bind-mounted configuration and SSL certificates.
labs/nextcloud-flow/docker-compose.yml:21-33labs/nextcloud-flow/nginx/nginx.conf:1-25nextcloud-aio-apache
Nextcloud web server
The Nextcloud Apache service, referenced in the nginx configuration as the upstream proxy target. It is not explicitly declared in the Compose file but is expected to be started by the AIO master container.
labs/nextcloud-flow/nginx/nginx.conf:12Windfall exploit scripts
exploit tools
Python scripts (windfall_afr.py and windfall_sqli.py) for path traversal and SQL injection exploits against Windmill/Flow. They are referenced in the README and defined as entry points in pyproject.toml.
labs/nextcloud-flow/README.md:37-55pyproject.toml:13-14Requirements and sequence described by the evidence
Prerequisites
- Add '127.0.0.1 localhost.local' to /etc/hosts on the host.
labs/nextcloud-flow/README.md:7-9 - Run 'docker compose up -d' to start the environment.
labs/nextcloud-flow/README.md:13-15 - Configure Nextcloud AIO via the web interface, including enabling Docker Socket Proxy.
labs/nextcloud-flow/README.md:19-26 - Install the Flow app in Nextcloud.
labs/nextcloud-flow/README.md:33-35
Evidence-described exercise path
- Start the Docker Compose environment.
labs/nextcloud-flow/README.md:13-15 - Configure Nextcloud AIO and start all containers.
labs/nextcloud-flow/README.md:19-26 - Install the Flow app in Nextcloud.
labs/nextcloud-flow/README.md:33-35 - Obtain the Flow container IP address.
labs/nextcloud-flow/README.md:39-40 - Run the path traversal exploit (windfall_afr.py) directly or via the Nextcloud proxy.
labs/nextcloud-flow/README.md:43-46 - Run the SQL injection exploit (windfall_sqli.py) directly or via the Nextcloud proxy.
labs/nextcloud-flow/README.md:48-55
Behaviors behind the stored safety assessment
No harmful behavior observed
The environment is a self-contained vulnerability lab. The Docker socket mount is read-only and required for AIO to manage containers. The nginx proxy and SSL certificates are for local access. No hidden, destructive, or external behaviors are visible.
labs/nextcloud-flow/docker-compose.yml:1-44labs/nextcloud-flow/README.md:1-71What the analysis did not establish
- The exploit scripts (windfall_afr.py, windfall_sqli.py) are not included in the evidence packet; their behavior is only described in the README.
- The Nextcloud AIO master container image is 'latest', which may change over time.
- The Docker socket mount, while read-only, could be a concern if the AIO container is compromised, but no evidence of such compromise is present.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.
GitHublabs/windmill-allinone/docker-compose.yml
Created Vuln labNo linked CVECompose · builds
Analysis
Environment assessment
A Docker Compose environment that builds a single container running Windmill 1.394.4 with an embedded PostgreSQL database, supervisor, and test data initialization scripts. The README describes it as a lab for testing path traversal vulnerabilities in Windmill.
labs/windmill-allinone/docker-compose.yml:1-21labs/windmill-allinone/Dockerfile:1-42README.md:1-490Lab assessment
Vulnerability labThe README explicitly states the environment is for testing path traversal and SQL injection vulnerabilities in Windmill, and the Dockerfile builds a specific vulnerable version (1.394.4) with test data containing fake credentials and secrets.
README.md:1-3labs/windmill-allinone/Dockerfile:5labs/windmill-allinone/init-testdata.sql:1-34Lab shapeCompose · builds
Services1
Compose manifests1
Dockerfiles1
Evidence & reasoningComponents · CVE assessment · exercise context · review boundaries
Services and files described by the evidence
windmill
vulnerable target applicationweb serverAPI server
A single service built from the Dockerfile, running Windmill 1.394.4 with an embedded PostgreSQL database. It exposes port 8000 internally, mapped to host port 8100. The container runs supervisor to manage PostgreSQL, Windmill, and a test data initialization script.
labs/windmill-allinone/docker-compose.yml:6-15labs/windmill-allinone/Dockerfile:5-42labs/windmill-allinone/supervisord.conf:1-36PostgreSQL
databasecredential store
PostgreSQL 15 installed and initialized inside the container, storing Windmill data and test credentials. It listens on localhost:5432 and is started by supervisor before Windmill.
labs/windmill-allinone/Dockerfile:9-24labs/windmill-allinone/supervisord.conf:7-14init-testdata
test data seeder
A shell script that waits for PostgreSQL and Windmill migrations, then executes init-testdata.sql to insert fake users, secrets, tokens, and resources into the database.
labs/windmill-allinone/init-testdata.sh:1-17labs/windmill-allinone/init-testdata.sql:1-34supervisor
process manager
Supervisor runs as PID 1, managing PostgreSQL, Windmill, and the init-testdata script. It ensures services start in the correct order and restarts them if they fail.
labs/windmill-allinone/supervisord.conf:1-36Requirements and sequence described by the evidence
Prerequisites
- Docker and Docker Compose must be installed to build and run the lab environment.
labs/windmill-allinone/docker-compose.yml:1-21 - The lab container must be built and started, which will automatically initialize the database and insert test data.
labs/windmill-allinone/Dockerfile:1-42labs/windmill-allinone/supervisord.conf:1-36 - The Windmill service must be accessible on the mapped host port (8100) to exercise the path traversal or SQL injection vulnerabilities.
labs/windmill-allinone/docker-compose.yml:9-10
Evidence-described exercise path
- Build and start the lab environment using 'docker compose up' from the labs/windmill-allinone directory.
labs/windmill-allinone/docker-compose.yml:1-21 - Wait for the container to become healthy (curl http://localhost:8000/api/version succeeds).
labs/windmill-allinone/docker-compose.yml:12-15 - Exploit the unauthenticated path traversal vulnerability (CVE-2026-29059) by sending a crafted request to the /api/w/X/jobs_u/get_log_file/ endpoint to read arbitrary files, such as /etc/passwd or /proc/self/environ.
README.md:155-160 - Leak credentials from the container (e.g., SUPERADMIN_SECRET from /proc/1/environ or PostgreSQL data files) to forge a JWT and achieve remote code execution as root inside the container.
README.md:161-180 - Optionally, use the authenticated SQL injection vulnerability (CVE-2026-23696) by creating a folder and injecting SQL via the addowner endpoint to escalate privileges and achieve RCE.
README.md:220-250
Behaviors behind the stored safety assessment
No harmful behavior observed
The lab is self-contained: it builds a vulnerable Windmill instance with fake test data, exposes only port 8100, and does not mount the Docker socket, run in privileged mode, or connect to external systems. All exploit behavior described in the README targets the lab's own container, and no hidden or unnecessary compromise is visible.
labs/windmill-allinone/docker-compose.yml:1-21labs/windmill-allinone/Dockerfile:1-42labs/windmill-allinone/init-testdata.sql:1-34What the analysis did not establish
- The packet does not include the exploit scripts (windfall_afr.py, windfall_sqli.py, windfall_nc_pivot.py) referenced in the README, so their exact behavior cannot be assessed.
- The Dockerfile uses a base image from ghcr.io/windmill-labs/windmill:1.394.4, which is not inspected; its contents are unknown.
- The README describes host escape via Docker socket, but the lab's docker-compose.yml does not mount the Docker socket, so that attack path is not part of this lab unit.
Model interpretation
This review is limited to the supplied lab evidence packet. It does not assert that the environment runs, reproduces a vulnerability, or is safe to execute. Contract: eip-docker-lab-analysis-v1.