What problem does it solve? A monitoring tool that collects findings, metrics, logs, and backups can quietly grow until it fills the very disk it watches. This Skill audits claude-watchman's own collected data under journal/ so the monitor never becomes the thing consuming the host's storage. ## Core Features & Use Cases - Footprint measurement: Sizes every artifact class (findings database, WAL/SHM sidecars, run.log, cost ledger, backups, monitor-state snapshots) via read-only retention_report and retention_total_mb calls. - Prunable-data enumeration: Counts terminal findings, old metrics/runs rows, oversized run.log, excess backups, and stale monitor-state snapshots without deleting anything. - Threshold-based journaling: Raises a single deduplicated finding (info/low/medium severity by growth) only when the footprint exceeds WATCHMAN_DATA_FOOTPRINT_WARN_MB or stale data exists, and records a trend metric. - Use Case: During a scheduled /watchman audit loop, the Skill detects that journal/ holds 71MB with 312 terminal findings older than 180 days, and journals a review-tier finding so the operator can prune it later via watchman fix. ## Quick Start Run the check-data-footprint skill during a watchman audit to measure the journal data footprint and report anything past its retention window.
