other changes

This commit is contained in:
Elizabeth W
2026-04-20 01:25:44 -06:00
parent 38ff2f4fde
commit 1036fce55e
11 changed files with 2035 additions and 171 deletions
+66 -1
View File
@@ -1,3 +1,68 @@
# agentguard-ci
A DevSecOps Argo Workflows pipeline to protect against AI coding agent hallucinations and supply chain attacks.
A DevSecOps Argo Workflows pipeline specifically designed to protect against AI coding agent hallucinations, supply chain attacks, and security misconfigurations in a homelab or solo-developer environment.
## 📖 The Problem
AI coding agents are highly productive "junior developers," but they lack intrinsic context. They frequently hallucinate dummy credentials, introduce insecure application logic, or pull in new, potentially typosquatted dependencies.
This pipeline acts as a strict, automated gatekeeper that prioritizes zero-noise alerting, allowing you to maintain high development velocity without compromising the security of your exposed homelab.
## 🏗️ Architecture & Features
This project deploys an **Argo ClusterWorkflowTemplate** that orchestrates a parallel security scanning matrix whenever code is pushed:
* **TruffleHog**: Verifies leaked API keys dynamically to prevent false-positives from AI hallucinations.
* **Semgrep**: Scans first-party application logic for vulnerabilities (e.g., SQLi, XSS).
* **Socket.dev**: Analyzes dependencies for supply chain attacks, malware, and typosquatting.
* **Pulumi CrossGuard**: Validates Infrastructure as Code against policy packs.
* **Syft + Grype**: Generates SBOMs and scans for container vulnerabilities scored via EPSS.
* **KICS**: Scans infrastructure misconfigurations.
* **DefectDojo & MinIO**: Uploads findings to a centralized ASPM dashboard and raw SARIF/JSON reports to S3-compatible storage.
* **Policy Enforcement**: Custom TypeScript logic automatically fails the build if any findings exceed your defined CVSS severity threshold.
For deep-dive architecture decisions, see the [Pipeline Overview ADR](docs/pipeline-overview.md) and [Secret Strategy ADR](docs/secret-strategy.md).
## 🚀 Prerequisites
Before installing the pipeline, ensure your Kubernetes cluster has the following installed:
* **Argo Workflows**
* **Infisical Kubernetes Operator** (for secret injection)
* **DefectDojo** (for vulnerability dashboards)
* **MinIO / S3** (for raw report storage)
You will also need API keys or tokens for: Socket.dev, Pulumi, AWS/MinIO, and DefectDojo.
## 🛠️ Installation
### 1. Build the Pipeline Tools Image
The pipeline relies on custom TypeScript logic (e.g., CVSS enforcement and API uploads). Build and push this image to your registry:
```bash
cd tools
docker build -t your-registry/agentguard-tools:latest .
docker push your-registry/agentguard-tools:latest
```
*(Make sure to update `clusterworkflowtemplate.yaml` with your custom image if you do not use `agentguard-tools:latest`)*
### 2. Configure Helm Values
Update `helm/values.yaml` (if applicable) and configure your Infisical integration:
```yaml
pipeline:
enabled: true
infisical:
workspaceSlug: "your-workspace-id"
projectSlug: "your-project-id"
```
### 3. Deploy via Helm
Install the pipeline and its associated resources to your cluster:
```bash
helm upgrade --install agentguard-ci ./helm -n argo
```
## 🔐 Secret Management Integration
To prevent hardcoded secrets in the pipeline, this project uses the **Infisical Kubernetes Operator**.
When you deploy the Helm chart, it creates an `InfisicalSecret` Custom Resource (`helm/templates/infisical-secret.yaml`). The Infisical Operator securely fetches your vault secrets (like `SOCKET_DEV_API_KEY` and `DEFECTDOJO_API_TOKEN`) and synchronizes them into a standard Kubernetes `Secret` named `amp-security-pipeline-secrets`.
The Argo Workflow then mounts this standard secret as environment variables inside the scanning containers, ensuring zero secret leakage in the Git repository.
+16
View File
@@ -0,0 +1,16 @@
import glob, re, os
files = glob.glob("helm/templates/scan-*.yaml") + glob.glob("helm/templates/upload-*.yaml") + ["helm/templates/enforce-policy.yaml"]
for f in files:
with open(f) as file:
content = file.read()
match = re.search(r'spec:\n templates:\n(.*)(?:{{- end }})', content, re.DOTALL)
if match:
template_content = match.group(1).strip()
# Extract the base name e.g. scan-kics
base_name = os.path.basename(f).replace('.yaml', '')
new_content = f'{{{{- define "template.{base_name}" }}}}\n{template_content}\n{{{{- end }}}}\n'
new_filename = os.path.join(os.path.dirname(f), f"_{base_name}.yaml")
with open(new_filename, "w") as out:
out.write(new_content)
os.remove(f)
+5 -1
View File
@@ -31,7 +31,7 @@ For solo personal projects, a complex CI/CD security pipeline is usually overkil
---
### The Chosen Solution: Dual-Layer Approach
### The Chosen Solution: Dual-Layer Approach + Infisical Runtime Injection
#### Layer 1: Gitleaks (The Local Guard)
* **Where:** Local developer machine (Pre-commit Hook).
@@ -41,6 +41,10 @@ For solo personal projects, a complex CI/CD security pipeline is usually overkil
* **Where:** GitHub Actions / CI Pipeline (Post-commit).
* **Why:** Uses active verification. If a secret slips past (via an AI agent pushing directly or a bypassed local hook), TruffleHog actively calls out to external APIs to verify if the key is live. By using the `--only-verified` flag, it guarantees zero false positives and only fails the pipeline if it proves a key is an active threat.
#### Layer 3: Infisical Operator (Pipeline Runtime Injection)
* **Where:** Inside the Kubernetes Cluster (via `InfisicalSecret` CRD).
* **Why:** The security pipeline itself requires numerous highly-privileged secrets (DefectDojo API tokens, AWS S3 keys, Pulumi access tokens, Socket.dev keys) to execute the scans and upload reports. We do not store these in GitOps. Instead, the Helm chart deploys an `InfisicalSecret` resource. The Infisical Kubernetes Operator authenticates with the central vault, pulls the secrets dynamically, and syncs them into a native Kubernetes `Secret` (`amp-security-pipeline-secrets`). The Argo Workflow containers then consume these safely at runtime as environment variables.
---
### Tradeoffs & Accepted Risks
+1 -1
View File
@@ -5,7 +5,7 @@ metadata:
data:
renovate.json: |
{
"extends": ["github>my-org/my-repo//renovate-preset"],
"extends": [{{ .Values.preset | quote }}],
"onboarding": false,
"platform": "github",
"repositories": {{ toJson .Values.repositories }}
+1
View File
@@ -4,4 +4,5 @@ image:
pullPolicy: IfNotPresent
schedule: "0 * * * *"
preset: "github>my-org/my-repo//renovate-preset"
repositories: []
+44 -80
View File
@@ -1,3 +1,4 @@
{{- if .Values.pipeline.enabled }}
apiVersion: argoproj.io/v1alpha1
kind: ClusterWorkflowTemplate
metadata:
@@ -47,21 +48,11 @@ spec:
value: "{{workflow.parameters.fail-on-cvss}}"
- name: upload-storage
dependencies:
- scan-trufflehog
- scan-semgrep
- scan-kics
- scan-socketdev
- scan-syft-grype
- scan-crossguard
- scanners
template: upload-storage
- name: upload-defectdojo
dependencies:
- scan-trufflehog
- scan-semgrep
- scan-kics
- scan-socketdev
- scan-syft-grype
- scan-crossguard
- scanners
template: upload-defectdojo
- name: enforce-policy
dependencies:
@@ -76,54 +67,6 @@ spec:
dependencies:
- scanners
template: sinks-and-enforcement
- name: scan-trufflehog
dependencies:
- clone
template: scan-trufflehog
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: scan-semgrep
dependencies:
- clone
template: scan-semgrep
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: scan-kics
dependencies:
- clone
template: scan-kics
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: scan-socketdev
dependencies:
- clone
template: scan-socketdev
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: scan-syft-grype
dependencies:
- clone
template: scan-syft-grype
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: scan-crossguard
dependencies:
- clone
template: scan-crossguard
arguments:
parameters:
- name: working-dir
value: "{{workflow.parameters.working-dir}}"
- name: clone-repo
inputs:
parameters:
@@ -148,39 +91,60 @@ spec:
tasks:
- name: trufflehog
template: scan-trufflehog
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: semgrep
template: scan-semgrep
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: kics
template: scan-kics
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: socketdev
template: scan-socketdev
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: syft-grype
template: scan-syft-grype
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: defectdojo
template: scan-crossguard
arguments:
parameters:
- name: working-dir
value: "{{inputs.parameters.working-dir}}"
- name: sinks-and-enforcement
container:
image: alpine:3.20
image: curlimages/curl:latest
command:
- sh
- -c
args:
- echo "stub: sinks and enforcement"
- name: scan-trufflehog
template: scan-trufflehog
- name: scan-semgrep
template: scan-semgrep
- name: scan-kics
template: scan-kics
- name: scan-socketdev
template: scan-socketdev
- name: scan-syft-grype
template: scan-syft-grype
- name: scan-crossguard
template: scan-crossguard
- name: upload-storage
template: upload-storage
- name: upload-defectdojo
template: upload-defectdojo
- name: enforce-policy
template: enforce-policy
- |
set -eu
echo "Pipeline complete. You can configure a webhook notification here."
if [ -n "${SLACK_WEBHOOK_URL:-}" ]; then
curl -X POST -H 'Content-type: application/json' --data '{"text":"Security Pipeline Finished"}' "${SLACK_WEBHOOK_URL}" || true
fi
{{ include "template.scan-syft-grype" . | indent 4 }}
{{ include "template.scan-socketdev" . | indent 4 }}
{{ include "template.scan-crossguard" . | indent 4 }}
{{ include "template.scan-semgrep" . | indent 4 }}
{{ include "template.scan-trufflehog" . | indent 4 }}
{{ include "template.scan-kics" . | indent 4 }}
{{ include "template.upload-defectdojo" . | indent 4 }}
{{ include "template.upload-storage" . | indent 4 }}
{{ include "template.enforce-policy" . | indent 4 }}
{{- end }}
-88
View File
@@ -1,88 +0,0 @@
{{- if .Values.pipeline.enabled }}
apiVersion: argoproj.io/v1alpha1
kind: ClusterWorkflowTemplate
metadata:
name: amp-security-pipeline-v1.0.0
spec:
templates:
- name: enforce-policy
inputs:
parameters:
- name: fail-on-cvss
container:
image: python:3.12-alpine
command:
- sh
- -c
args:
- |
set -eu
python - <<'PY'
import json
import os
import pathlib
import sys
threshold = float(os.environ["FAIL_ON_CVSS"])
reports_dir = pathlib.Path("/workspace/reports")
findings = []
for report in sorted(reports_dir.iterdir()):
if not report.is_file():
continue
text = report.read_text(errors="ignore")
if report.suffix == ".sarif":
try:
data = json.loads(text)
except json.JSONDecodeError:
continue
for run in data.get("runs", []):
for result in run.get("results", []):
for fix in result.get("properties", {}).get("security-severity", []):
pass
for level in result.get("properties", {}).values():
pass
for prop in [result.get("properties", {}), result.get("taxa", [])]:
pass
for region in result.get("locations", []):
pass
sev = result.get("properties", {}).get("security-severity")
if sev is None:
continue
try:
score = float(sev)
except (TypeError, ValueError):
continue
if score >= threshold:
findings.append((report.name, score))
elif report.suffix == ".json":
try:
data = json.loads(text)
except json.JSONDecodeError:
continue
if isinstance(data, dict):
for item in data.get("findings", data.get("vulnerabilities", [])):
score = item.get("cvss") or item.get("score")
if score is None:
continue
try:
score = float(score)
except (TypeError, ValueError):
continue
if score >= threshold:
findings.append((report.name, score))
if findings:
for name, score in findings:
print(f"{name}: CVSS {score} >= {threshold}", file=sys.stderr)
raise SystemExit(1)
print(f"No findings met or exceeded CVSS {threshold}")
PY
env:
- name: FAIL_ON_CVSS
value: "{{inputs.parameters.fail-on-cvss}}"
volumeMounts:
- name: workspace
mountPath: /workspace
{{- end }}
+14
View File
@@ -0,0 +1,14 @@
FROM node:20-alpine
WORKDIR /app
COPY package.json package-lock.json ./
RUN npm ci
COPY tsconfig.json ./
COPY src ./src
RUN npm run build
# The default command isn't strictly necessary as Argo will override it
CMD ["node", "/app/dist/enforce-policy.js"]
+1853
View File
File diff suppressed because it is too large Load Diff
+21
View File
@@ -0,0 +1,21 @@
{
"name": "tools",
"version": "1.0.0",
"description": "",
"main": "index.js",
"type": "module",
"scripts": {
"test": "vitest run",
"build": "tsc"
},
"keywords": [],
"author": "",
"license": "ISC",
"type": "commonjs",
"devDependencies": {
"@types/node": "^25.6.0",
"tsx": "^4.21.0",
"typescript": "^6.0.3",
"vitest": "^4.1.4"
}
}
+14
View File
@@ -0,0 +1,14 @@
{
"compilerOptions": {
"target": "ES2022",
"module": "NodeNext",
"moduleResolution": "NodeNext",
"outDir": "./dist",
"rootDir": "./src",
"strict": true,
"esModuleInterop": true,
"skipLibCheck": true,
"forceConsistentCasingInFileNames": true
},
"include": ["src/**/*"]
}