The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill handles incident descriptions and monitoring alerts that may contain untrusted external data.
Ingestion points: scripts/incident_triage.py accepts incident descriptions via the --description command-line argument, and scripts/handle_alerts.py processes alert titles and descriptions from monitoring sources.
Boundary markers: The skill does not implement delimiters or instructions to prevent the agent from interpreting data within these fields as commands or instructions.
Capability inventory: The automation scripts are capable of writing reports to the local file system and sending notifications to external webhooks and APIs.
Sanitization: Input text is interpolated into JSON payloads and Markdown reports without explicit sanitization or filtering logic.
[DATA_EXFILTRATION]: The skill is configured to transmit incident metadata and descriptions to external services.
Evidence: scripts/handle_alerts.py contains methods _send_slack_notification and _send_pagerduty_notification which use the requests library to send data to external endpoints.
Context: These network operations target well-known incident management and communication platforms (Slack and PagerDuty) as part of the intended response workflow.

incident-responder