The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes a script that fetches data from an external incident management system, creating an attack surface where malicious data could influence agent behavior.
Ingestion points: The check_thehive_cases and calculate_ir_metrics functions in scripts/agent.py retrieve case details from an external API.
Boundary markers: None identified in the script or documentation to delimit external data from instructions.
Capability inventory: The script can perform network POST requests and write to local files (ir_playbook_output.json).
Sanitization: External content is parsed as JSON but not explicitly sanitized or validated before being included in generated reports.
[DATA_EXFILTRATION]: The skill performs network operations using the requests library in scripts/agent.py to communicate with external APIs based on user-provided URLs. While this is consistent with the stated purpose, it represents a network capability to non-whitelisted domains.

building-incident-response-playbook