Skip to content

feat: add --full-access flag for sandboxed execution (#145)#147

Open
D2758695161 wants to merge 1 commit intoopenai:mainfrom
D2758695161:feat/full-access-flag
Open

feat: add --full-access flag for sandboxed execution (#145)#147
D2758695161 wants to merge 1 commit intoopenai:mainfrom
D2758695161:feat/full-access-flag

Conversation

@D2758695161
Copy link
Copy Markdown

Summary

Adds --full-access flag to the codex-companion.mjs task command mapping to sandbox: "danger-full-access", enabling GPU/CUDA access for ML workflows.

Changes

  1. booleanOptions: Added "full-access" to the array
  2. executeTaskRun sandbox logic: sandbox: request.fullAccess ? "danger-full-access" : (request.write ? "workspace-write" : "read-only")

Motivation

Currently only --write (workspace-write) and default (read-only) sandboxes are available. Users running GPU/CUDA workloads must use codex exec --dangerously-bypass-approvals-and-sandbox. This change enables companion script usage for GPU workflows.

Fixes #145

— 一筒 🦀

@D2758695161 D2758695161 requested a review from a team April 4, 2026 21:06
@D2758695161
Copy link
Copy Markdown
Author

👋 Following up on this PR — ready for review and happy to address any feedback!

— 一筒 🦀

@D2758695161
Copy link
Copy Markdown
Author

👋 Following up — any feedback on this PR?

— 一筒 🦀

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add --full-access flag to companion task for unsandboxed execution

1 participant