[WIP] Fix failing GitHub Actions workflow test

[Copilot]

Fix LLM Gateway Support Flags for Codex and Claude Engines

Problem Analysis

The GitHub Actions test suite is failing with two test failures in TestSupportsLLMGateway and TestValidateStrictFirewall_LLMGatewaySupport:

Failing Tests:

• TestSupportsLLMGateway/Codex_engine_supports_LLM_gateway - expects Codex to support LLM gateway (true)
• TestSupportsLLMGateway/Claude_engine_does_not_support_LLM_gateway - expects Claude to NOT support LLM gateway (false)
• TestValidateStrictFirewall_LLMGatewaySupport/codex_engine_with_LLM_gateway_rejects_sandbox.agent:_false_in_strict_mode - expects Codex to have LLM gateway support

Current State:

• pkg/workflow/codex_engine.go:44 - supportsLLMGateway: false ❌
• pkg/workflow/claude_engine.go:33 - supportsLLMGateway: true ❌

Expected State (from tests):

• Codex engine should support LLM gateway (true) ✅
• Claude engine should NOT support LLM gateway (false) ✅

Changes Made

• Analyze failing test logs and identify root cause
• Update Codex engine to set supportsLLMGateway: true
• Update Claude engine to set supportsLLMGateway: false
• Update enable_api_proxy_test.go to match new expectations
• Update docker_api_proxy_test.go to match new expectations
• Run selective tests to verify the fix
• Run full test suite to ensure no regressions
• Review and commit changes

Original prompt

Fix the failing GitHub Actions workflow test
Analyze the workflow logs, identify the root cause of the failure, and implement a fix.
Job ID: 63600190114
Job URL: https://github.com/github/gh-aw/actions/runs/22009453612/job/63600190114

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[Copilot]

Pull request overview

This PR fixes failing test suite by correcting the LLM gateway support flags for Codex and Claude engines. The tests were expecting Codex to support LLM gateway while Claude does not, but the implementation had these flags reversed.

Changes:

• Codex engine now correctly reports LLM gateway support (supportsLLMGateway: true), enabling api-proxy integration
• Claude engine now correctly reports no LLM gateway support (supportsLLMGateway: false), skipping api-proxy setup
• All related tests updated to match the corrected engine capabilities

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
pkg/workflow/codex_engine.go	Sets supportsLLMGateway: true to enable api-proxy sidecar for LLM API proxying through AWF firewall
pkg/workflow/claude_engine.go	Sets supportsLLMGateway: false and aligns comment formatting
pkg/workflow/enable_api_proxy_test.go	Updates test expectations: Claude should NOT include --enable-api-proxy flag, Codex should include it
pkg/workflow/docker_api_proxy_test.go	Updates test expectations: Claude should NOT include api-proxy image, Codex should include it

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The comment in pkg/workflow/docker.go (lines 109-111) describes the api-proxy container as supporting both "Port 10000: OpenAI API proxy (for Codex)" and "Port 10001: Anthropic API proxy (for Claude)". However, with this change, only Codex will use the LLM gateway (supportsLLMGateway: true), while Claude will not (supportsLLMGateway: false). This means the api-proxy container will only be deployed when using Codex, making the Port 10001 reference potentially misleading. Consider either: (1) updating the comment to clarify that Port 10001 is for future use, (2) updating the comment to remove the Claude reference, or (3) if both engines should support LLM gateway, reconsidering whether Claude should also have supportsLLMGateway: true.

Suggested change

	supportsLLMGateway: false, // Claude does not support LLM gateway
	supportsLLMGateway: true, // Claude supports LLM gateway