[CI] Fix Claude GitHub Actions setup (#1401)

## Summary

Fix Claude GitHub Actions setup added in #1400 

## Test plan

- [ ] Comment `/claude review` on this PR and verify the deep-review job
runs (this is also serving as the smoke-test for the workflows merged in
#1400)
- [ ] Comment `@claude any thoughts on this change?` and verify the
interactive job runs
- [ ] Check that both jobs respect the new 10-minute cap if they ever
stall

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Chores**
  * Updated CI/CD workflow timeouts for job execution management.
* Disabled experimental AI beta features in CI runs to ensure consistent
behavior.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Keval Morabia <28916987+kevalmorabia97@users.noreply.github.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: kevalmorabia97

.github/workflows/claude.yml

@@ -52,6 +52,7 @@ jobs:
         )
       )
     runs-on: ubuntu-latest
+    timeout-minutes: 10
     permissions:
       contents: read
       pull-requests: write
@@ -67,6 +68,13 @@ jobs:
         uses: anthropics/claude-code-action@v1
         env:
           ANTHROPIC_BASE_URL: ${{ secrets.ANTHROPIC_BASE_URL }}
+          # NVIDIA inference proxy (LiteLLM-based) rejects two fields
+          # the Claude Code SDK sends by default. Set per NVIDIA/OSMO's
+          # workflow which has hit and solved both issues:
+          #   - `context_management` → disable experimental betas
+          #   - `cache_control.ephemeral.scope` → disable prompt caching
+          CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: "1"
+          DISABLE_PROMPT_CACHING: "1"
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
           claude_args: |

.github/workflows/claude_review.yml

@@ -23,6 +23,7 @@ jobs:
       contains(github.event.comment.body, '/claude review') &&
       contains(fromJson('["OWNER", "MEMBER", "COLLABORATOR"]'), github.event.comment.author_association)
     runs-on: ubuntu-latest
+    timeout-minutes: 10
     permissions:
       contents: read
       pull-requests: write
@@ -59,6 +60,13 @@ jobs:
         uses: anthropics/claude-code-action@v1
         env:
           ANTHROPIC_BASE_URL: ${{ secrets.ANTHROPIC_BASE_URL }}
+          # NVIDIA inference proxy (LiteLLM-based) rejects two fields
+          # the Claude Code SDK sends by default. Set per NVIDIA/OSMO's
+          # workflow which has hit and solved both issues:
+          #   - `context_management` → disable experimental betas
+          #   - `cache_control.ephemeral.scope` → disable prompt caching
+          CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: "1"
+          DISABLE_PROMPT_CACHING: "1"
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
           trigger_phrase: "/claude review"