The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided JSON configuration files which are used to generate reports and dashboards, representing a potential surface for indirect prompt injection. • Ingestion points: The scripts alert_optimizer.py, dashboard_generator.py, and slo_designer.py load local JSON files via the load_alert_config and load_service_definition methods. • Boundary markers: No specific boundary markers or 'ignore' instructions are present to delimit processed data. • Capability inventory: The scripts possess file-write capabilities through methods like export_analysis and export_specification. • Sanitization: The _generate_html_report method in alert_optimizer.py interpolates user-supplied alert names into an HTML template without escaping, which is a best-practice violation.
[SAFE]: The skill includes sample assets in the assets/ directory that contain non-functional placeholder information for demonstration purposes. • Evidence: assets/sample_alerts.json contains a stub Slack webhook URL (https://hooks.slack.com/services/warnings) and placeholder routing keys (pager-key-critical) that do not represent active credentials.

observability-designer