pipeguard-cli 0.1.1

Pre-commit security scanner for GitHub Actions workflows

PipeGuard

Catch GitHub Actions security issues before they reach your runners.

Pre-commit security scanner for GitHub Actions workflows. Catches supply-chain risks, unpinned actions, known CVEs, and secret leaks — before you push.

---

Demo

The recording shows a scan of a real workflow file with multiple security issues. PipeGuard detects them all in under a second — no API key, no network call, fully offline:

• No top-level permissions: block — GitHub grants write access to all scopes by default
• tj-actions/changed-files@v35 and reviewdog/action-setup@v1 match known CVEs in the bundled database
• All actions are pinned to tags instead of commit SHAs — a supply-chain risk
• echo ${{ secrets.DEPLOY_TOKEN }} leaks a secret value to the workflow log
• 8398a7/action-slack is a third-party action from an unverified publisher

---

PipeGuard vs. the alternatives

Feature	PipeGuard	actionlint	StepSecurity	act
SHA-pinning check	✅	❌	✅	❌
CVE database (offline)	✅	❌	✅ (online)	❌
Permissions analysis	✅	⚠️ syntax only	✅	❌
Secrets-leak detection	✅	❌	✅ (runtime)	❌
Supply-chain audit	✅	❌	✅	❌
Syntax / type checks	✅ via actionlint	✅	❌	❌
Run workflows locally	❌	❌	❌	✅
Runtime hardening	❌	❌	✅	❌
Pre-commit hook	✅	✅	❌	❌
SARIF output	✅	✅	❌	❌
No API key required	✅	✅	❌	✅
Fully offline	✅	✅	❌	⚠️ needs images
Open source	✅	✅	✅	✅
Free	✅ core	✅	⚠️ freemium	✅

PipeGuard fills the gap between authoring and execution: static security analysis, offline, before you push, without any external service.

---

Installation

Prerequisites: Python 3.11+

pip install pipeguard-cli

Install actionlint (required for syntax checks, optional for other checks):

# Linux
curl -fsSL https://github.com/rhysd/actionlint/releases/latest/download/actionlint_linux_amd64.tar.gz \
  | tar xz actionlint && sudo mv actionlint /usr/local/bin/

# macOS
brew install actionlint

# Ubuntu/Debian (snap)
sudo snap install actionlint

---

Usage

pipeguard scan [PATH]

PATH can be a single workflow file or a directory. If omitted, defaults to .github/workflows.

Examples

# Scan entire .github/workflows directory (default)
pipeguard scan

# Scan a specific directory
pipeguard scan .github/workflows/

# Scan a single file
pipeguard scan .github/workflows/ci.yml

# JSON output (for scripts / CI pipelines)
pipeguard scan --format json

# SARIF output (for IDE integration, GitHub Code Scanning)
pipeguard scan --format sarif

# Use a custom config file
pipeguard scan --config path/to/.pipeguard.yml

When scanning a directory, pipeguard prints a per-file header and a summary at the end:

Scanning: .github/workflows/ci.yml
  ✓ No issues found

Scanning: .github/workflows/deploy.yml
  ╭─────────┬──────────────────┬──────────────────────────┬──────────────╮
  │ Severity│ Rule             │ Location                 │ Message      │
  ├─────────┼──────────────────┼──────────────────────────┼──────────────┤
  │ error   │ cve-cve-2025-... │ deploy.yml:12            │ Action ...   │
  │ error   │ sha-pinning      │ deploy.yml:15            │ Action ...   │
  ╰─────────┴──────────────────┴──────────────────────────┴──────────────╯

Scanned 2 file(s) — 2 error(s), 1 warning(s), 3 info(s) total.

Exit codes

Code	Meaning
0	No errors or warnings
1	One or more errors or warnings found

Info-level findings (action inventory) do not affect the exit code.

---

Security checks

Errors & Warnings

Rule	Severity	Description
sha-pinning	error	Action pinned to a tag or branch instead of a full commit SHA — supply-chain risk (cf. CVE-2025-30066)
sha-pinning-reusable	error	Reusable workflow called with a tag or branch ref instead of a full commit SHA — same supply-chain risk as unpinned actions
cve-<id>	error	Action matches a known CVE in the local database (offline, no API)
supply-chain	warning	Third-party action from an unverified publisher — suppress per publisher or action via .pipeguard.yml
secrets-leak	error	Secret value echoed or logged in a run: step
secrets-leak-debug	error / warning	set -x or set -o xtrace in a run: step — shell debug mode prints every expanded command; error when secrets are in env scope, warning otherwise
permissions-missing	error	No permissions: block — GitHub grants write access to most scopes by default
permissions-write-all	error	permissions: write-all at top-level or per-job
permissions-excessive	warning	Sensitive scope (e.g. secrets, workflows) set to write
permissions-id-token-unused	warning	id-token: write is set but no OIDC-consuming action is present — allows minting cloud credentials unnecessarily
permissions-invalid	error	Unknown permission level (not read, write, or none)
pull-request-target	warning	pull_request_target trigger runs with base-repo write permissions and secret access — even for fork PRs
pull-request-target-pwn	error	pull_request_target combined with checkout of the PR head ref — allows untrusted fork code to run with secrets (Pwn Request)
actionlint	error	Syntax and type errors (requires actionlint)

Info

Rule	Severity	Description
action-inventory	info	Inventory of all third-party actions used — ref(s) and occurrence count

---

CVE database

PipeGuard ships with a built-in offline CVE database — no API key, no network call during scan. The database is updated daily via NVD and shipped with each new weekly release.

[!NOTE] Last updated: 2026-03-12

CVE	Action	Description
CVE-2025-30066	tj-actions/changed-files	tj-actions/changed-files was compromised in a supply-chain attack. All tag-based refs are potentially affected — pin to a known-good SHA.
CVE-2025-30154	reviewdog/action-setup	reviewdog/action-setup was compromised in the same supply-chain campaign as tj-actions/changed-files. Pin to a vetted SHA.
CVE-2023-49291	tj-actions/branch-names	tj-actions/branch-names improperly references github.event.pull_request.head.ref and github.head_ref in a run step, allowing a specially crafted branch name to execute arbitrary code and steal secrets or abuse GITHUB_TOKEN permissions.

SHA-pinned refs are not flagged — another reason to always pin to a commit SHA.

To get the latest CVEs, upgrade to the newest release:

pip install --upgrade pipeguard-cli

---

Configuration

PipeGuard looks for a .pipeguard.yml (or .pipeguard.yaml, pipeguard.yml, pipeguard.yaml) in the current directory and its parents.

Suppress supply-chain warnings for trusted publishers

If you trust a specific publisher or action, add them to the allowlist:

# .pipeguard.yml
trusted_publishers:
  - my-org           # suppresses all warnings for my-org/*
  - some-vendor      # suppresses all warnings for some-vendor/*

trusted_actions:
  - other-org/specific-action  # suppresses warning for this exact action only

• trusted_publishers matches as a prefix — every action from my-org/* is considered trusted.
• trusted_actions requires an exact match on the action name (without the @ref).
• A trailing / in publisher names is optional; PipeGuard normalises it automatically.

The built-in allowlist (GitHub, AWS, Azure, Google, Docker, HashiCorp, etc.) is always active and cannot be removed.

---

CI integration

Add pipeguard as a pre-push check or CI step:

# .github/workflows/pipeguard.yml
- name: Scan workflows
  run: |
    pip install pipeguard-cli
    pipeguard scan .github/workflows/

Or as a pre-commit hook (.pre-commit-config.yaml):

repos:
  - repo: local
    hooks:
      - id: pipeguard
        name: PipeGuard workflow scan
        entry: pipeguard scan
        language: python
        additional_dependencies: [pipeguard-cli]
        files: \.github/workflows/.*\.ya?ml$

---

Output formats

Terminal (default) — human-readable table with rule, location, and fix suggestions.

JSON — machine-readable, one object per finding:

[
  {
    "rule": "cve-cve-2025-30066",
    "severity": "error",
    "message": "Action 'tj-actions/changed-files@v35' is affected by CVE-2025-30066: ...",
    "file": ".github/workflows/ci.yml",
    "line": 12,
    "fix_suggestion": "Immediately pin 'tj-actions/changed-files' to a verified safe SHA."
  }
]

SARIF — compatible with GitHub Code Scanning.

---

License

Apache 2.0 — see LICENSE.

Status

In active development. All Free-tier scanners are functional. Planned: auto-fix PRs (Pro), Maintainer Trust Score (Pro), cloud sandbox testing, IDE plugin.

Free-tier checks: SHA-pinning, CVE database, permissions audit (incl. id-token: write without OIDC), secrets-leak detection, action inventory, actionlint syntax checks.