The Agent Skills Directory

[Command Execution]: The skill relies on execution of shell commands via gcloud, bq, and curl to retrieve infrastructure logs and metrics. This is a primary function of the skill for cloud resource management and follows standard administrative patterns.
[Dynamic Execution]: In the cost estimation reference, the skill generates and executes Python code strings using python3 -c. This is used as an in-memory calculator to parse JSON data from Monitoring API responses and perform tiered pricing calculations, avoiding local file system persistence.
[Indirect Prompt Injection Surface]: The skill processes untrusted data sourced from cloud logs (e.g., VPC Flow Logs, Threat Logs) which could potentially contain attacker-controlled content.
Ingestion points: Data enters through list_log_entries, execute_sql, and API calls in references/vpc-flow-logs-cost-estimation.md and references/vpc-flow-analysis.md.
Boundary markers: The skill uses LIMIT clauses and schema verification steps to scope data ingestion.
Capability inventory: The agent has access to run_shell_command, BigQuery, and Cloud Logging tools.
Sanitization: The instructions utilize structured data handling like JSON_VALUE in SQL and json.dumps in Python to manage external content.

google-cloud-networking-observability