S.01Infrastructure benchmark

CI/CD and infrastructure pentest benchmark

CI/CD and infrastructure pentest benchmark across Jenkins, Redis, PostgreSQL, Terraform and Vault, real results only: Jenkins script console RCE, Redis unauthenticated exploitation, PostgreSQL COPY PROGRAM RCE, Terraform tfstate secrets exposed and a Docker socket container escape, each proven with a working exploit.

Star Darkmoon on GitHub All benchmarks

135

infrastructure vulnerabilities

CI/CD and data-layer labs

proved with a real exploit

Darkmoon, the open source autonomous AI penetration testing tool, found 135 infrastructure vulnerabilities across 6 CI/CD and data-layer labs and proved 47 of them with a real exploit, from anonymous access to root on the host.

S.03Attack chains
What each run proved

Every tile opens the long-form write-up. The chip is the report's own findings total; the line above the title is its exploited count and severity split.

41 findings

8 exploited · 15C · 15H · 9M · 2L

Vault + registry + Docker socket

Guessed root token, image-layer secrets, and a Docker-socket container escape reading the host /etc/shadow.

34 findings

16 exploited · 21C · 5H · 8M

Terraform + AWS + Ansible

An exposed terraform.tfstate and Ansible inventory chained to an AdministratorAccess CI key.

26 findings

2 exploited · 4C · 8H · 10M · 2L · 2I

GitLab CE 19.2.1

An admin PAT with api and sudo scopes, an unmasked AWS secret in CI/CD variables, and open signup.

22 findings

13 exploited · 6C · 10H · 6M

PostgreSQL 16 + MySQL 5.6

PostgreSQL COPY TO PROGRAM RCE, pg_shadow and mysql.user hashes, and live session tokens.

9 findings

5 exploited · 3C · 5H · 1M

Redis 7.4.10 (unauth)

Unauthenticated access and the admin session token extracted; the RDB-write RCE honestly demoted (Redis 7.x).

3 findings

3 exploited · 3C

Jenkins 2.541.3 (security off)

Anonymous HTTP to script-console RCE to the master encryption key. All three findings exploited.

S.04Results
Infrastructure runs, finding by finding

Each row links to its long-form write-up and to the raw report in the research corpus.

Lab / target	Findings	Severity	Exploited	Model	Evidence
Vault + registry + Docker socketcamp_20260801_2bd90d3f	41	15C15H9M2L	8	claude-opus-4-6	Write-up Report
Terraform + AWS + Ansiblecamp_20260801_b1b96939	34	21C5H8M	16	claude-opus-4-6	Write-up Report
GitLab CE 19.2.1camp_20260801_a719641d	26	4C8H10M2L2I	2	claude-opus-4-6	Write-up Report
PostgreSQL 16 + MySQL 5.6camp_20260801_c0151524	22	6C10H6M	13	claude-opus-4-6	Write-up Report
Redis 7.4.10 (unauth)camp_20260801_96be38b9	9	3C5H1M	5	claude-opus-4-6	Write-up Report
Jenkins 2.541.3 (security off)camp_20260801_b6ad197d	3	3C	3	claude-opus-4-6	Write-up Report

Disclaimer
Darkmoon's own benchmark on public infrastructure labs.

The offensive runs are produced by the open source Darkmoon CLI; the web dashboard and the remediation-to-PR loop are paid Pro. Raw reports live in the darkmoon-research corpus and the results feed the Darkmoon-Benchmarks leaderboard.

darkmoon-research corpus Darkmoon-Benchmarks leaderboard

S.06More benchmarks
Keep exploring the benchmarks

Cloud benchmark

AWS and Azure cloud penetration testing benchmark

AWS, Azure and GCP identity, storage and metadata chains.

IoT benchmark

IoT firmware penetration testing benchmark

OWASP IoTGoat, static firmware and a live appliance.

Web benchmark

Web app pentest benchmark: OWASP Juice Shop

OWASP Juice Shop, black-box, six-campaign escalation.

Pro

Automated remediation benchmark (Pro): 42 of 57 fix PRs

57 findings, 57 pull requests, 42 demonstrated end to end on OWASP Juice Shop; the 15 excluded cases are disclosed in full.

S.07FAQ
Infrastructure benchmark questions

What does the CI/CD and infrastructure pentest benchmark cover?

Six autonomous runs against Jenkins, Redis, PostgreSQL and MySQL, Terraform with AWS and Ansible, HashiCorp Vault with a container registry and the Docker socket, and GitLab. Highlights include Jenkins script console RCE, Redis unauthenticated exploitation, PostgreSQL COPY PROGRAM RCE, Terraform tfstate secrets exposed and a Docker socket container escape.

How many infrastructure vulnerabilities did Darkmoon find?

Darkmoon, the open source autonomous AI penetration testing tool, found 135 infrastructure and CI/CD vulnerabilities across 6 labs and proved 47 of them with a real exploit. The Redis run also honestly demoted the classic RDB-write RCE, because Redis 7.x mitigates it.

Was the infrastructure benchmark run on a local model?

This lab wave ran on claude-opus-4-6 through the open source Darkmoon CLI, which can also run on a local model (Ollama or llama.cpp). The Privacy Gateway tokenizes real values so the model works on placeholders. The web dashboard and the remediation-to-PR loop are paid Pro capabilities.

S.08Next
Run these checks in your own pipeline

Open source, self hosted and local first. Wire Darkmoon into CI/CD and gate the build on exploited findings. A star helps other teams find it.