[Bug]: repo-monitor ignores AUTOMATION_MODEL profile

Bug Description

skills/github-repo-monitor/scripts/main.py builds its conversation payload from the agent server's active profile and never reads AUTOMATION_MODEL, so a repo-monitor automation runs on whatever profile is active in the UI instead of its configured one. (Same bug as #428, which has no type label so the readiness bot cannot evaluate it; the maintainer narrowed the remaining scope to repo-monitor in #428.)

Actual Behavior

Reproducible without any server:

python -m pytest tests/test_github_repo_monitor.py -q -k profile_resolves
# fails on main: no request to /api/profiles/<AUTOMATION_MODEL> is ever made,
# the active profile's llm is forwarded unchanged

With AUTOMATION_MODEL=gpt-5.6-sol set, _get_agent_dict returns the active profile's llm (openai/gpt-5.5) instead of the configured profile.

Expected Behavior

When AUTOMATION_MODEL is set, resolve that profile via GET /api/profiles/{name} (plaintext secrets) and forward its config with usage_id: profile:<name>; fall back to the active profile when unset or on 404.

Acceptance Criteria

• AUTOMATION_MODEL set + profile exists → forwarded llm is the profile config with usage_id: profile:<name>.
• AUTOMATION_MODEL set + profile 404 → falls back to the active profile.
• AUTOMATION_MODEL unset → active profile, no profile request.
• New tests/test_github_repo_monitor.py passes.

Start with skills/github-repo-monitor/scripts/main.py, especially _get_agent_dict, and run python -m pytest tests/test_github_repo_monitor.py -q -k profile_resolves. Verify the configured AUTOMATION_MODEL profile is requested and forwarded with its usage_id, while unset or 404 cases retain the active profile behavior; then run the full named test file.