OIDC App token exchange returning 500 Internal Server Error since Mar 18 12:00 UTC

## Problem

All `claude-code-action` workflows using `claude_code_oauth_token` have been failing since approximately **2026-03-18 12:00 UTC** with:

```
App token exchange failed: 500 Internal Server Error - Internal server error
Retrying in 5 seconds...
Attempt 1 failed: Internal server error
...
Operation failed after 3 attempts
```

This affects all three retry attempts. The action never reaches Claude execution.

## Impact

- All PR-triggered workflows using `claude_code_oauth_token` are broken
- Issue comment workflows (`@claude`) are also affected
- This has been ongoing for **7+ hours** with no recovery

## Reproduction

Any workflow using:
```yaml
- uses: anthropics/claude-code-action@v1
with:
claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
```

Fails immediately at OIDC token exchange.

## Environment

- Repository: `oviva-ag/claude-code-plugin`
- Action version: `anthropics/claude-code-action@v1` (SHA: 1b422b3517b51140e4484faab676c5e68b914866)
- Coincides with Anthropic status page incident: "Elevated errors on Opus 4.6" reported at 12:30 UTC

## Example failed runs

- https://github.com/oviva-ag/claude-code-plugin/actions/runs/23263904494
- https://github.com/oviva-ag/claude-code-plugin/actions/runs/23247230397
- https://github.com/oviva-ag/claude-code-plugin/actions/runs/23244292276

Start by running the workflow shown in the reproduction section and inspect the linked failed runs around the OIDC token exchange. Confirm whether the exchange still returns HTTP 500 and compare the behavior with the cited Anthropic status incident. Done means workflows using claude_code_oauth_token complete token exchange and reach Claude execution.

Written by the indexing model from the issue text.