The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill documents numerous bash commands for the gcloud and bq CLIs. This is the intended primary purpose of the skill and no arbitrary or suspicious command injection patterns were found.
[CREDENTIALS_UNSAFE] (SAFE): While the skill describes authentication procedures (service accounts, auth login), it uses safe placeholders like key.json and PROJECT_ID. No actual private keys or tokens are hardcoded.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill features commands that ingest external, potentially untrusted data into the agent's context.
Ingestion points: gcloud logging read, gcloud logging tail, and gcloud storage cp gs://bucket/file.txt ./ (reading logs or files from cloud storage).
Boundary markers: Absent in the command templates.
Capability inventory: Extensive GCP resource management including IAM policy modification, VM creation, and secret access.
Sanitization: None provided within the skill text; the agent relies on its own output filtering.

gcloud