The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: Evaluated ingestion surfaces for potential injection attacks.
Ingestion points: The skill ingests raw input parameters $0 (scan type) and $1 (watchlist) directly from user space via variables.
Boundary markers: No validation rules or explicit sanitization constraints are provided to boundaries or user arguments before string interpolation into path variables.
Capability inventory: Uses Write tool to create scanners/{scan_type}/ and {scan_type}_scanner.py scripts natively based on variable configurations, accompanied by Bash execution capabilities.
Sanitization: Lacks string filtering or layout validation mechanisms to clean input text before formatting it directly into system commands and file paths.
[DYNAMIC_EXECUTION]: Script paths and target directories (scanners/{scan_type}/ and {scan_type}_scanner.py) are computed dynamically on-demand from runtime execution arguments.

indicator-scanner