How should trusted Symphony deployments enable .git writes and localhost runtime validation?

[woody-korean]

On March 6, 2026, I ran Symphony Elixir against a Git-backed app repo and hit the same blocker across repeated retries.

Local workflow posture:

codex:
  approval_policy: never
  thread_sandbox: workspace-write
  turn_sandbox_policy:
    type: workspaceWrite

This matches the documented safe defaults / supported values in the repo docs.

Observed behavior:

• the agent could edit normal repo files in the per-issue workspace
• the agent could not complete .git writes needed for branch / commit / publish flow
• the agent could not complete localhost launch-path validation in-session
• the issue stayed in In Review, and retries kept finding the same partial state (src/App.tsx and src/App.css edits only)

That makes the current unattended profile feel like an implementation-only sandbox, not an end-to-end "land the PR safely" profile.

Question:
What is the intended trusted-environment setup when you want Symphony to actually finish the full repo workflow, including:

1. branch / commit / push / PR work that touches .git
2. localhost runtime validation / launch-path QA

Is the intended answer one of these?

• run with danger-full-access / dangerFullAccess
• keep Codex sandboxed and move publish / localhost validation into external hooks or host automation
• use a different app-server capability / sandbox profile than the one shown in the current examples

A short docs section or example "trusted publish profile" would help a lot. Right now it is easy to copy a safe profile that can edit files but still dead-end on the last mile.