The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes security alert data from external tools such as Wiz and SentinelOne, creating a potential surface for indirect prompt injection.
Ingestion points: Alert details, asset environment, and data sensitivity levels are ingested via the scripts/security_triage.py script's command-line arguments or JSON input.
Boundary markers: The skill uses a structured JSON schema and clear command-line parameters to separate user-provided data from script logic.
Capability inventory: The skill executes a local Python script and has the capability to write results to a user-defined file path.
Sanitization: The triage script includes robust input validation, such as clamping CVSS scores to the 0.0-10.0 range and verifying asset/data categories against predefined allowed values.
[EXTERNAL_DOWNLOADS]: The skill references several external resources for regulatory guidelines and tool documentation.
The skill fetches release information from MITRE's official GitHub repository and links to government portals (gov.il) and established security vendor documentation (docs.wiz.io, docs.snyk.io). These are recognized as trusted or well-known services and do not represent a security risk.
[COMMAND_EXECUTION]: The skill includes a local triage script (scripts/security_triage.py) designed for execution via a standard Python interpreter to automate alert scoring. The script operates on provided inputs without spawning unsafe subprocesses or executing arbitrary strings.

israeli-cybersecurity-ops