The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content from project configuration files that could influence agent behavior.
Ingestion points: The skill reads contents from Dockerfile, cloudbuild.yaml, nginx.conf, and vite.config.ts during the audit process.
Boundary markers: No explicit delimiters or instructions are used to separate the content of external files from the agent's instructions.
Capability inventory: The skill has significant capabilities, including executing shell commands via Bash (using docker, gcloud, and kubectl tools) and making network requests via WebFetch.
Sanitization: Data from external files is analyzed via direct reading and grep without specific sanitization to filter out potential prompt injection attempts.
[COMMAND_EXECUTION]: The skill utilizes Bash to run validation scripts and standard deployment tools. It performs architecture checks, audits build arguments, and verifies deployment status using gcloud, kubectl, and docker commands.

gcp-deploy-guardian