Observability
Security checks across malware telemetry and agentic risk
Overview
This appears to be a purpose-aligned observability skill, but it can log/export agent activity and run a local monitoring dashboard, so users should avoid putting secrets in telemetry and keep the dashboard private.
Before installing, review the Node package and lockfile, run the dashboard only on a trusted local machine, avoid placing secrets in trace/log/alert metadata, and configure any webhook destinations carefully. The provided evidence supports a benign observability tool, not credential theft or malicious behavior.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
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.
Anyone who can reach the dashboard port may be able to view operational logs and metrics.
The skill intentionally starts a local REST dashboard with endpoints for status, metrics, logs, Prometheus metrics, and alerts. This matches the purpose, but it is still an exposed local service.
启动 Dashboard 服务器: npm start / node src/dashboard.js ... 访问 http://localhost:3001 ... `/api/logs` GET 最近日志
Run it only on a trusted machine, keep the port bound to localhost or firewalled, and add authentication before exposing it remotely.
It may be harder to verify the origin and intended build/run process of the bundled Node project.
The registry metadata gives limited provenance and no install spec, while the included project files document npm-based setup and execution. This is a reviewability/provenance note rather than evidence of malicious behavior.
Source: unknown; Homepage: none; No install spec — this is an instruction-only skill.
Review package.json and package-lock.json, install in an isolated project, and avoid running unreviewed updates from unknown sources.
Sensitive data passed into trace metadata, errors, or contexts could end up in logs and be visible through log export/dashboard features.
The system logs caller-supplied trace metadata and error context. For an observability tool this is expected, but those fields may contain prompts, file paths, tool parameters, stack traces, or other sensitive details.
this.logger.info(`[TRACE START] ${operation}`, { traceId, spanId, ...meta }); ... this.logger.error('Error occurred', { error: error.message, stack: error.stack, ...context });Avoid putting secrets or full prompts in metadata, add redaction for tokens/PII, and control access to log files and the dashboard.
Alert messages and rule metadata may leave the local environment if a webhook channel is configured.
Alert rules can use a webhook channel that sends alert and rule details to a configured URL. This is a normal alerting feature, but it creates an external data flow when enabled.
this.channels = config.channels || ['console']; this.webhookUrl = config.webhookUrl || null; ... const payload = { alert: alert, rule: rule.toJSON(), timestamp: new Date().toISOString() };Use only trusted HTTPS webhook destinations and avoid including secrets in alert names, messages, thresholds, or metadata.