The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze infrastructure-as-code and configuration files (Terraform, Deployment Manager, IAM policies) which may contain untrusted content. This creates a surface for indirect prompt injection where malicious text in the data could attempt to override the agent's behavior.
Ingestion points: SKILL.md identifies file types like **/*.tf, **/*.yaml, and **/*.json as targets for analysis.
Boundary markers: The skill includes a dedicated 'Prompt Injection Safety Notice' section that explicitly instructs the agent to treat string values and comments as data, not instructions.
Capability inventory: The skill is restricted to read-only tools (Read, Grep, Glob) and lacks capabilities for file modification, command execution, or network communication.
Sanitization: The skill relies on prompt-level instructions to maintain the distinction between data and instructions during processing.

gcp-review