The Agent Skills Directory

Command Execution: The skill utilizes the gcloud CLI tool to create cloud controls and frameworks. This involves executing system commands to manage cloud infrastructure, which is the intended purpose of the skill.
Indirect Prompt Injection: The skill is designed to process data from external files (/tmp/author_handoff.json) and web search results to define compliance rules. This ingestion process is a potential surface for indirect prompt injection. Evidence: Ingestion points include local temporary files and search results; boundary markers consist of instructions to treat data as passive; capabilities include gcloud write operations; sanitization relies on specific prompt instructions to ignore embedded commands.
Dynamic Execution: The skill invokes Python scripts (db_resources_loader.py and db_output_match_analyzer.py) located in a related directory to analyze existing resource coverage. This represents dynamic execution of local logic as part of its workflow.
External Resource Reference: The skill refers to official Google Cloud documentation and consoles for configuration and guidance. These references are used to assist the user in managing their environment.

assuredworkloads-frameworks-author