Skip to content

Failure Flags

Each failure is intended to be flagged with four attributes:

  • Category — which task family the failure occurred in.
  • Signal type — the specific failure mode (the declared what went wrong).
  • Severity — how serious the flag is, on the same flag-driven scale used across the ecosystem (critical vs non-critical).
  • Recommended remediation — the suggested next step to repair the gap.

Under the planned design, critical failures (security · safety) would route to immediate operator review, while non-critical failures would aggregate into the SwarmFixer repair queue once that component exists. As with the live DefendableCloud referee, a flag is a rulebook outcome — never a quality opinion or a judge-model verdict.


🐝 Operator-grade · books and records · to the shed.