# Playbook: **When to reach for this:** **What it observes:** Every signal below should trace to one of them. --- ## §1: — e.g. "Is the work queue backing up?" **Inspect:** ``` ``` **Baseline healthy:** Give the reader a number to compare against, not a vibe. **What looks wrong → what it means → what to do:** - **** (e.g. "queue depth climbing, oldest job > 10 min") → → . - **** → → . --- ## §2: <" just fired — what now?"> **Inspect:** ``` ``` **Baseline healthy:** **What looks wrong → what it means → what to do:** - **** → → . - **Deadlock/wedge note:** --- ## §N: > Keep adding entries until every signal the substrate emits has one. The test of completeness: pick any > alert or metric this substrate can produce — is there an entry that tells a fresh operator what healthy > is, what wrong is, and what to do? If not, the substrate's observability is incomplete. --- ## Cross-references - The design doc's observability section (the signals this playbook interprets). - The substrate's runbook / recovery tools. - Related playbooks for adjacent substrates.