gcp
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION] (LOW): The skill provides a reference for using the
Bashtool to executegcloud,gsutil, andbqcommands. These tools are powerful and can modify infrastructure permissions (e.g.,gsutil iam ch allUsers:objectViewerwhich makes a bucket public) or deploy unauthenticated services (--allow-unauthenticated). - [DATA_EXPOSURE] (LOW): The documentation includes standard authentication procedures (
gcloud auth login) and credential retrieval (gcloud container clusters get-credentials) which involve the creation and management of sensitive tokens on the local environment. - [INDIRECT_PROMPT_INJECTION] (LOW):
- Ingestion points: Untrusted data can enter the agent's context through several commands described in
SKILL.md, includinggsutil cp,bq query, andpubsub subscriptions pull. - Boundary markers: None identified in the provided command reference or SDK snippets.
- Capability inventory: The skill allows the
Bashtool, providing the agent with broad execution capabilities over the host and network (SKILL.md). - Sanitization: There are no examples or instructions regarding the sanitization of data retrieved from GCP services before it is processed or used in further commands.
Audit Metadata