Skip to main content

Source: docs/manual/incident-triage.md

This page is generated by site/scripts/sync-manual-docs.mjs.

Incident Triage Engine

Cruvero includes a specialist incident-triage engine centered on SonarTriageWorkflow for root-cause generation, low-risk fix planning, simulation, and tracked remediation output.

Source: internal/specialists/sonartriage/*, internal/specialists/workflows.go, cmd/worker/main.go

Workflow Scope​

SonarTriageWorkflow executes a deterministic sequence of specialist activities:

  1. Fetch critical/blocker issue candidates.
  2. Build root-cause analysis hints.
  3. Generate minimal remediation plans.
  4. Simulate fix safety and blast-radius constraints.
  5. Submit issue/PR skeleton outputs when safe.
  6. Apply compensation and escalation paths when blocked.

The workflow supports one-shot and continuous modes.

Core Safety Mechanics​

  • Plan hashing and simulation hashing for repeatability.
  • Risk score gating with escalation reason tracking.
  • Compensation hints for rollback paths.
  • Cleanup ledger for idempotent operation handling.
  • Blackboard append events for reflection/escalation telemetry.

Incident Communication Integrations​

The workflow can invoke communication tools through MCP tool runners.

Current repository state includes simulated messaging tools for demos/tests:

  • sim_teams_post (simulated Microsoft Teams posting)
  • additional simulated incident tools (sim_cost_regression, sim_git_pr, etc.)

Production Slack/Teams posting behavior depends on enabled real MCP tools in the deployment environment.

Registration and Runtime Wiring​

  • Specialist workflows and activities are registered in worker startup (cmd/worker/main.go).
  • Canonical workflow type: SonarTriageWorkflow.
  • Agent profile integration is wired through specialist profiles and prompt templates.