Artifact Attestations public beta

Create a tamper-proof papertrail for anything you build on Actions

Artifact Attestations lets you sign builds in GitHub Actions, capturing provenance information about the artifact and making it verifiable from anywhere. There are no keys or PKI to manage, and verification happens with the GitHub CLI tool. The solution is based on Sigstore, an open source project that simplifies signing for software artifacts.

To add provenance to a GitHub Actions workflow, you just need to invoke the new attest-build-provenance Action with the path to an artifact. Here’s a simple example:

permissions:
  id-token: write
  contents: read
  attestations: write

#
# (build your artifact)
#

- name: Generate artifact attestation
  uses: actions/attest-build-provenance@v1
  with:
    subject-path: 'PATH/TO/ARTIFACT'

Then verify it with the CLI tool:

gh attestation verify PATH/TO/ARTIFACT -o myorganization

To learn more check out the blog and join the discussion in the GitHub Community.

Developers of GitHub Apps can simplify their application by using the client ID for both OAuth flows and the installation token flow.

To date, GitHub Apps have had two different IDs to manage – the application ID and the client ID. The application ID was only used to mint a JWT, subsequently used to fetch an installation token. The client ID is used with the OAuth flow to sign in users and request installations. These two values equally identify the application and the question of which one to use where caused unnecessary developer friction. You can now use the client ID in the place of the application ID when minting JWTs.

The application ID is not being deprecated at this time, nor are their plans to remove it. However, compatibility with future features will rely on use of the client ID, so updating is recommended.

The specific change allowed here is that when minting the JWT that proves your app is in posession of an application’s private key, you can use the client ID for the iss claim. Note that application IDs are ints, while client IDs are strings, if using a typed language.

require 'openssl'
require 'jwt'  # https://rubygems.org/gems/jwt

# Private key contents
private_pem = File.read("YOUR_PATH_TO_PEM")
private_key = OpenSSL::PKey::RSA.new(private_pem)

# Generate the JWT
 payload = {
 # issued at time, 60 seconds in the past to allow for clock drift
  iat: Time.now.to_i - 60,
  # JWT expiration time (10 minute maximum)
  exp: Time.now.to_i + (10 * 60),
--- # GitHub App's App ID
--- iss: "12345"
+++ # GitHub App's Client ID
+++ iss: "Iv23f8doAlphaNumer1c"
}

jwt = JWT.encode(payload, private_key, "RS256")
puts jwt

Note that Octokit still expects the use of the App ID in its setup – the Octokit SDK will be updated in the future to support use of the client ID.

You can find the client ID for your application in its settings page:

A screenshot of an app's settings, showing both the client ID and the application ID

Client IDs and application IDs are not secrets, and are expected to be visible to the end user – you do not need to change how you handle your IDs when making this update.

For more information about minting JWTs to get an installation token, see ‘Generating a JWT for a GitHub App’.

See more