quantumquirkxyz Contract
- Input: problem or task defined by the skill body.
- Output: Markdown artifact or structured result with completion criteria met.
- Side effects: none (design/review/documentation only unless explicitly stated).
- Dependencies: none (self-contained unless linked to other skills).
- Stop condition: all process steps completed; artifact saved; criteria checked.
- Risk: low.
- Boundary: produces reasoning or documentation artifacts; does not modify external systems unless explicitly instructed.
Monitoring and Alerting
Use this skill when a project needs to stay visible after deploy. It should decide what to watch, what to alert on, who responds, and what a human should inspect first when something goes wrong. observability defines the signal vocabulary; this skill turns the relevant signals into operational decisions.
Contract
- Input: monitoring brief, runtime surface, and failure modes.
- Output: a monitoring plan, alert priorities, diagnostic signals, and response ownership.
- Scope: design the alert surface, not the monitoring stack implementation.
- Rule: prioritize user-visible failures and operational regressions.
- Rule: avoid alert spam that does not change a maintainer's decision.
- Rule: connect alerts to the seam where the system actually degrades.
- Rule: every paging alert needs a threshold or condition, owner, severity, response target, and runbook or first diagnostic action.
- Rule: prefer symptoms and service-level indicators for paging; use causes as supporting diagnostics unless they require immediate action.
- Rule: define retention, sampling, cardinality, and redaction constraints for signals that may contain user or secret data.
- Rule: state how deploy markers and release identity correlate with alerts so regressions can be attributed.
Steps
- Identify the user-visible failure modes, service-level indicators, and operational risks that matter most.
- Map each failure mode to signals, dashboards, deploy markers, and alerts with explicit thresholds.
- Assign severity, owner, response target, escalation, and a runbook or first diagnostic action.
- Check signal cost and safety: retention, sampling, cardinality, redaction, and false-positive risk.
- Define the smallest useful alert set and the evidence needed to tune it after real incidents.
Completion criteria
- the top failure modes are named
- alert conditions, severity, and ownership are named
- the initial diagnostic path and escalation are explicit
- signal privacy and operational cost constraints are addressed