Bug type

Behavior bug (incorrect output/state without crash)

Summary

Embedded and subagent runs ignore the parent agent's per-agent model override and silently fall back to the global default, causing unexpected errors when the global default provider is unavailable.

Steps to reproduce

1. Configure a per-agent model override in agents.list[] that differs from agents.defaults.model.primary

2. Send a message that triggers the agent to spawn an embedded or subagent run

3. Observe that the embedded run uses agents.defaults.model.primary instead of the parent agent's configured model

Expected behavior

Embedded and subagent runs should inherit the parent agent's agents.list[].model.primary when spawned by that agent.

Actual behavior

Embedded and subagent runs resolve their model from agents.defaults.model.primary, ignoring the parent agent's per-agent override.

OpenClaw version

2026.3.13

Operating system

Ubuntu 24.04

Install method

npm global

Model

openai-codex gpt-5.4

Provider / routing chain

Gateway (moltbot.service → /opt/moltbot/openclaw/openclaw.mjs) Model resolver → resolves to agents.defaults.model.primary (openai-codex/gpt-5.4) Auth profile → openai-codex:default (OAuth mode) Provider → OpenAI Codex API (OAuth) → rate limited Fallback resolver → agents.defaults.model.fallbacks[0] (minimax/MiniMax-M2.5) Provider → https://api.minimax.io/anthropic → 404 model not found Fallback resolver → deepseek/deepseek-reasoner Provider → https://api.deepseek.com/v1 (API key auth, openai-completions API) → succeeds No external proxy or model router — it's direct gateway-to-provider with the built-in failover chain. The ~10 minute latency is accumulated across steps 4–7 (cooldown windows between fallback attempts).

Config file / key location

No response

Additional provider/model setup details

No response

Logs, screenshots, and evidence

[gateway] agent model: openai-codex/gpt-5.4

[agent/embedded] embedded run agent end: isError=true model=gpt-5.4 provider=openai-codex error=⚠️ API rate limit reached
[agent/embedded] embedded run failover decision: decision=fallback_model reason=rate_limit provider=openai-codex/gpt-5.4
[model-fallback/decision] decision=candidate_failed requested=openai-codex/gpt-5.4 reason=rate_limit next=deepseek/deepseek-reasoner
[model-fallback/decision] decision=candidate_succeeded requested=openai-codex/gpt-5.4 candidate=deepseek/deepseek-reasoner

Impact and severity

No response

Additional information

No response