Signals overview

Signals is the unified queue of things that might need improvement. Every 👎, every error, every verification block, every silent regression — all lands here. Signals aren’t tasks yet. They’re candidates. You (or the system) decide which ones are real, which ones cluster together, and which ones get promoted into an Improvement.

​

Where signals come from

​

The signal lifecycle

1. Detected — a source (feedback, error, verification, etc.) fires.
2. Enriched — Nora adds context: the trace, the user, the intent, any prior similar signals.
3. Clustered — similar signals join together (see Clusters).
4. Reviewed — you confirm, mute, or edit the signal. Muting excludes false alarms.
5. Promoted — confirmed signals (or clusters) become Improvements in the improvement flow.
6. Resolved — the underlying issue is fixed and the signal closes.

Signals never delete themselves — they close as resolved or mute as false alarms. The audit trail matters.

​

The queue view

Signals is a sortable, filterable list. Columns:

• Severity — high / medium / low. Auto-computed from source type, impact, and frequency.
• Source — which channel fired (feedback, verification, etc.).
• Cluster — cluster this signal belongs to, if any.
• First seen — when the pattern first appeared.
• Occurrences — how many times this pattern (or cluster) has fired.
• Status — new / reviewing / muted / promoted / resolved.

Common views: “New this week”, “Unresolved high-severity”, “Clusters with 5+ occurrences”.

​

Working through signals

Recommended cadence:

• Daily — glance at high-severity news signals. Confirm or mute.
• Weekly — review clusters. Promote to Improvements.
• After a Flow change — check regressions specifically (see Regressions).

​

Where to go next