monitor-check

/monitor check

Run a metric health check on one or all monitors.

Instructions

Step 1: Determine scope

• If argument provided (monitor ID): check that specific monitor
• If --all flag: check all active monitors
• If no argument: show list and ask user which to check, or offer "check all"

Step 2: Read monitor configuration

For each monitor to check:

1. Read the monitor file from .analysis/metrics/monitors/{ID}_{slug}.md
2. Read the linked metric definition from .analysis/metrics/definitions/
3. Note: check cadence, comparison basis, thresholds, auto-segment setting

Step 3: Gather data

Ask the user for the current metric value(s).

Three data collection paths:

• MCP query: If MCP is configured in config.md, suggest running a query
• Manual input: "What's the current value of {metric}?"
• File/screenshot: "Paste or upload the latest data"

If auto-segment is enabled:

• Also ask for segment-level values (or suggest MCP query to get breakdown)

Step 4: Evaluate

Compare the current value against the monitor's configuration.

First check (no history)? If the Check History table is empty (this is the first check):

• Record the current value as the baseline — skip comparison-based evaluation
• Evaluate against absolute thresholds only (healthy range from the metric definition), if defined
• If no absolute thresholds available, mark status as 📊 Baseline recorded
• Tell the user: "This is the first check. Recording {value} as your baseline. Meaningful comparisons will begin on the next check."
• Proceed directly to Step 7 (Update monitor file) — no alerts on the first check

Normal check (history exists):

Comparison calculation:

Basis	Formula
WoW	(current - last_week) / last_week × 100
MoM	(current - last_month) / last_month × 100
YoY	(current - same_period_last_year) / same_period_last_year × 100
Rolling avg	current vs {N}-day rolling average
Fixed target	current vs target