Anomaly Watcher
PassAudited by ClawScan on May 10, 2026.
Overview
Anomaly Watcher appears to be a coherent local security-monitoring skill, but it persistently logs broad agent activity that may include sensitive details.
This skill looks purpose-aligned for local anomaly monitoring. Before installing, confirm you are comfortable with persistent .security logs of agent activity, decide what data should be excluded, and make sure you know how to disable monitoring and manage log retention.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Local monitoring logs may retain sensitive activity history and could influence future baseline calculations.
The skill persists arbitrary metric details and session identifiers into local JSONL logs. This is expected for anomaly monitoring, but those details could include sensitive prompt, tool, path, or session context depending on what is passed in.
details: metric.details || {}, session: metric.session || 'unknown' ... fs.appendFileSync(metricsFile, JSON.stringify(entry) + '\n');Use this only in workspaces where persistent security logs are acceptable, and configure or review what details are recorded, how long they are retained, and who can read the .security directory.
If configured as described, the skill may continue recording agent activity across many interactions.
The skill is intended to operate continuously through hooks. That is consistent with its monitoring purpose, but it means the behavior is broader than a one-time user command when enabled.
Always-on monitoring via hooks: `PostToolUse` — log every tool invocation; `UserPromptSubmit` — log input patterns; `PostSkillExecution` — log skill results
Enable the hook-based mode only if you want continuous monitoring, and make sure there is a clear way to disable it and clear or archive old logs.