AI Docs Bundle
Join Our Community Slack Contact

AI Documentation Security

Learn about the security measures and compilation process for Chainguard's AI documentation bundles
  3 min read

Overview

Chainguard’s AI documentation bundles are compiled with multiple security measures to ensure developers can trust the content they’re using with AI coding assistants. This page details our security practices and compilation process.

Security Measures

1. Automated Security Scanning

Every compilation runs through multiple security checks:

  • Secret Detection: We scan for API keys, tokens, and other sensitive data
  • Pattern Matching: Common secret patterns are automatically redacted
  • Bundle Size Limit: The build fails if the compiled bundle exceeds 50 MB
  • Extension Filtering: Only .md, .html, .json, and .yaml files are processed

2. Cryptographic Signatures

All documentation bundles are signed using Sigstore/Cosign:

  • Keyless Signing: Using OIDC identity verification
  • Transparency Log: All signatures recorded in Rekor
  • Certificate Chain: Full certificate provided for verification
  • Multiple Signatures: Both individual files and bundles are signed

3. Content Integrity

We ensure content hasn’t been tampered with:

  • SHA-256 Checksums: For all files in the bundle
  • Signed Checksums: The checksum file itself is signed
  • Build Provenance: GitHub Actions workflow attestations
  • Container Signing: Images signed by immutable digest with Cosign

Compilation Process

Source Repositories

Documentation is compiled from the chainguard-dev/edu repository:

  1. Documentation pages: every page published on this site
  2. Dockerfile Converter mappings: the package and image mappings from chainguard-dev/dfc, which a nightly job copies into the edu repository

Build Environment

  • GitHub Actions: Secure, ephemeral build environment
  • Resource Limits: CPU and memory constraints enforced
  • Restricted Egress: Network access limited to required endpoints via StepSecurity Harden Runner
  • Minimal Permissions: Only required repository access

What Gets Filtered

During compilation, we automatically remove:

  • Environment variables and secrets
  • Internal URLs and endpoints
  • Base64 encoded data blocks
  • Private key materials
  • Authentication tokens

Example patterns we redact:

  • api_key=...
  • password=...
  • -----BEGIN PRIVATE KEY-----
  • GitHub tokens (ghp_, ghs_)

Verification Guide

Container Image Verification

Verify the container image signature before pulling documentation:

cosign verify ghcr.io/chainguard-dev/ai-docs:latest \
  --certificate-identity-regexp ".*github.com/chainguard-dev/edu.*" \
  --certificate-oidc-issuer https://token.actions.githubusercontent.com

Build Frequency

  • Scheduled Builds: Nightly at 2 AM UTC
  • On-Demand: Triggered by documentation changes and by changes to the compilation scripts
  • Container Distribution: Updated container pushed to GHCR on each build

Security Reporting

If you discover a security issue:

  1. Do NOT open a public issue
  2. Email security@chainguard.dev
  3. Include:
    • Description of the issue
    • Steps to reproduce
    • Potential impact

FAQ

Why are some sections marked [REDACTED]?

This indicates our security scanner detected potentially sensitive information and removed it to protect our systems and users.

Can I build the bundle myself?

Yes! The compilation scripts are open source:

git clone https://github.com/chainguard-dev/edu
cd edu
python3 scripts/compile_docs.py

The compiler reads only the edu repository, so a clone is all it needs. It writes the bundle to static/downloads/chainguard-complete-docs.md.

How do I verify the build logs?

Build logs are public on GitHub Actions:

What if verification fails?

  1. Ensure you have the latest version of cosign
  2. Check your internet connection (for transparency log verification)
  3. Try downloading the files again
  4. Report persistent issues to support@chainguard.dev

Additional Resources

Last updated: 2026-09-30 13:19