The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various gcloud CLI commands to manage project configurations, authenticate accounts, and switch between project profiles.
[CREDENTIALS_UNSAFE]: The skill provides instructions for using service account JSON keys stored at ~/.gcp/{SA_KEY_FILE}.json and utilizes gcloud auth application-default print-access-token which outputs sensitive access tokens to the console.
[PROMPT_INJECTION]: Exhibits an indirect prompt injection surface as it interpolates user-supplied data (GCP Project IDs and Profile Names) directly into shell commands and Python code.
Ingestion points: User-provided strings for {PROJECT_ID} and {PROFILE_NAME} in SKILL.md.
Boundary markers: Absent; values are interpolated directly into command strings.
Capability inventory: Subprocess execution via gcloud commands and Python script execution via the google-cloud-bigquery library.
Sanitization: No explicit sanitization or validation of the project ID or profile name format is mentioned.

bigquery-auth