Multi_agent_v2 fails against Bedrock Mantle Endpoint

[jwalkins2017]

codex-security version 0.1.20

When I am trying to use Bedrock Mantle as my model provider

codex-secruity scan test-repo/ --provider amazon-bedrock --model openai.gpt-5.6-terra --headless --verbose

I end up seeing an error of

codex-security: debug: runtime.cleanup.started
codex-security: debug: runtime.cleanup.completed
codex-security: debug: scan.failed classification="unknown" partial_output=true
{"error":{"code":"validation_error","message":"invalid request body: Invalid 'input': value did not match any expected variant","param":null,"type":"invalid_request_error"}}
Partial output was kept at /state/codex-security/scans/test-repo/codex-security-test-repo-ijcmCp

This is related to openai/codex#39804, I have found a workaround of using

codex-secruity scan test-repo/ --provider amazon-bedrock --model openai.gpt-5.6-terra --headless --verbose --codex features.multi_agent_v2.max_concurrent_threads_per_session=1

[Hughhhhcoder]

The reported reproduction uses Codex Security 0.1.20. The current main/0.1.25 source has since moved the Deep Scan executor into the canonical plugin source in #675, and its worker config now explicitly sets features.multi_agent_v2.enabled=false while preserving the worker capacity settings.

I cannot confirm from a local run whether that fully covers the Bedrock Mantle path, so I have not opened a speculative configuration PR. If this still reproduces on the current release, could you please retest and share the current effective worker configuration (with credentials redacted), subagent count, and the exact failing request/model response? That will distinguish the 0.1.20 runtime path from a still-current provider incompatibility.