gcp-pipeline-orchestration
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill performs extensive shell command execution using
gcloudandbqtools to initialize, validate, deploy, and trigger orchestration pipelines on Google Cloud Platform.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by reading parameters (such as pipeline IDs and environment names) from local YAML files and interpolating them into shell commands.\n - Ingestion points: The agent reads
deployment.yamland orchestration pipeline definitions from the workspace root (e.g.,SKILL.mdSteps 1, 3, 4, 6, and 7).\n - Boundary markers: No specific boundary markers or "ignore instructions" warnings are utilized when processing these files.\n
- Capability inventory: The skill has access to powerful CLI tools (
gcloud,bq) and the ability to execute a local Python script (scripts/trigger/airflow_trigger.py).\n - Sanitization: Although the provided schema reference includes regex validation for certain fields and the
validatecommand is used, the skill does not explicitly instruct the agent to sanitize all file-derived inputs before executing them in the shell.\n- [EXTERNAL_DOWNLOADS]: Thescripts/trigger/airflow_trigger.pyscript performs network operations to interact with Google Cloud Composer APIs and Airflow instances. These operations target well-known Google Cloud infrastructure.
Audit Metadata