Repository navigation
Conversation
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| } | ||
|
|
||
| export function gatewayKeyFingerprint(key: string) { | ||
| return createHash("sha256").update(key).digest("hex"); |
erulkey
changed the title
This was referenced Oct 14, 2026
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Split into smaller PRs for review, in dependency order:
The completed stack preserves this PR’s implementation. This original PR remains open as a reference.
Sandbox applications and coding agents currently need separately supplied model credentials. Add
--aito request a Gateway key for each running session using the caller’s Vercel login and inject it through the external firewall. The managed secret never reaches the CLI or VM.On the universal image, the first command configures and launches Codex; bare
--aiprepares all four bundled agents without launching one. Custom images keep their application and agent configuration unchanged, including on reconnect. Applications must target Gateway using the placeholderAI_GATEWAY_API_KEYsupplied in the VM. Existing shortcuts without--airetain their authentication behavior.--aiignores host and saved keys. Reusing your own key requires--ai --use-gateway-key, which reads hostAI_GATEWAY_API_KEYor a key saved bysandbox gateway setup. Missing or rejected credentials fail without silently switching modes. Setup prints the credential mode to stderr unless--silentis set. Caller-owned keys remain bound to their original sandbox and are not revoked on stop.Managed issuance requires vercel/api#98063 and the team rollout flag. This backend is undeployed, so the default managed flow is not ready for production use. Agent shortcuts require the universal image; runtime and snapshot overrides remain unsupported with
--ai.Validation
300 CLI tests pass with sequential file execution, along with CLI typechecking and CLI/SDK builds. The built CLI, tested against a simulated API, requests managed credentials despite a host key being present, uses that key only with the explicit flag, and preserves API rejection without fallback. Tests also cover stdout/stderr, silent output, failed setup and legacy reconnect. README, generated help and the Sandbox skill are updated.
Earlier personal-Pro live tests verified Gateway inference on universal and non-universal Node images, preservation of custom-image configuration and reconnect. These used caller-owned keys before explicit selection was added; they did not verify managed issuance or cleanup. An earlier Codex launch completed inference and a file action after command approval, but its bundled terminal isolation hit a bubblewrap capabilities error, so unattended tool execution remains unverified.
Release still needs deployed managed lifecycle checks, all-four-agent inference/tool actions and the staging event/permission/authentication prerequisites documented in the backend handoff. Full-day use and startup benchmarks are untested. Parallel suites have intermittently failed an existing mocked credential-concurrency test; this change leaves that code untouched.