incidents-to-agent

Recurring incident failures trigger new agent candidate generation when three or more self-review failures occur within the same task type. The system applies filters based on agent eligibility and ensures no existing pending decisions before producing stubs.

Escalation to create a new agent candidate is triggered when three or more self-review failures occur within the same task type. The workflow applies agent eligibility filters and verifies no pending decisions exist before producing candidate stubs.

The workflow supports both automatic and manual triggers for escalation. It processes incident trails to generate agent candidates via Python scripts, implementing cleanup actions upon candidate approval or decline.

Decision flow management for proposed agent candidates uses batch AskUserQuestion interactions to reflect decisions. Upon approval, the system updates related indexes and performs file cleanup; upon decline, it removes the generated stubs.

Agent candidate generation from incident trails requires three or more self-review failures within the same task type. It enforces an agent eligibility filter and ensures no existing pending decisions are present before generating new candidate stubs.