Stabilize none runtime in agent-os

[kai-linux]

Goal

Fix repeated environment failures in the agent dispatcher to improve from the current 0% success baseline.

Success Criteria

• Identify the dominant environment failure mode affecting none in agent-os
• Land one bounded fix or preflight guardrail that prevents recurrence
• Improve weekly success rate snapshot beyond the current 0/17 baseline

Constraints

• Prefer minimal fixes over comprehensive refactoring
• Investigate root cause before adding workarounds
• Add diagnostic logging to detect future recurrence

Next Steps

• Inspect recent none failures in agent-os grouped by runner or environment cause
• Implement the smallest fix that prevents the dominant failure mode
• Add logging to verify the fix prevents the same path from failing again

Evidence

• agent_remediation:none:unknown:environment (agent_scorer): none completed 0 of 17 tasks in the last 7 days in agent-os, below the 60% threshold; likely cause: environment.

Reasoning

None at 0% success due to environment instability is blocking all downstream agent fallbacks; fixing dispatcher stability unblocks the entire queue.