Skip to content

[Optimus] [Bug] Release-engineer agent snapshot had invalid model claude-opus-4.6-1m — caused 5 consecutive release-gate crashes #551

@cloga

Description

@cloga

Problem

The release-gate cron suffered 5 consecutive immediate-crash failures starting 2026-03-31T04:30Z through 2026-04-01T04:30Z.

Root cause diagnosed by patrol Run #314:

  • The T2 agent snapshot .optimus/agents/release-engineer_0b24e54c.md contained model: claude-opus-4.6-1m
  • User's ~/.optimus/config/available-agents.json only has claude-opus-4.6 and gpt-5.4 in available_models for claude-code engine
  • worker-spawner.ts model pre-flight check (!allowedModels.includes(activeModel)) threw immediately — Task PID died within 1s, no output produced

Immediate fix applied by patrol Run #314:

  • Changed snapshot field from claude-opus-4.6-1mclaude-opus-4.6 in release-engineer_0b24e54c.md

Investigation Needed

  1. How did claude-opus-4.6-1m appear in the snapshot? The snapshot was auto-created on 2026-03-31T00:31Z. What model was requested or inferred at creation time? Is there a bug in how model is precipitated into T2 snapshots?
  2. Should worker-spawner auto-fallback to first available model instead of hard-crashing on mismatch?
  3. Should cron failures trigger alerting? 5 crashes with no auto-escalation is a gap.

Labels Requested

  • P1, bug, release-gate, agent-reliability

🤖 Created by patrol-manager via Optimus Spartan Swarm

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions