The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill requests and utilizes the Bash(cmd:*) tool, which allows for unrestricted execution of arbitrary shell commands. While the intended use case is deployment automation and scaffolding, there are no constraints on what commands can be executed.
[PROMPT_INJECTION] (HIGH): The skill is highly susceptible to Indirect Prompt Injection (Category 8).
Ingestion points: Untrusted data enters the context via user-provided "requirements" and "agent goals" in Step 1 of the instructions, as well as via existing code when "Refactoring" (Step 4).
Boundary markers: No delimiters or "ignore embedded instructions" warnings are present to separate instructions from the data being processed.
Capability inventory: The agent has full file-system access (Read, Write, Edit) and full shell access (Bash).
Sanitization: There is no mention of sanitizing or validating user input or existing code before it influences the behavior of the Bash or Write tools.
[EXTERNAL_DOWNLOADS] (LOW): The skill references Google Cloud documentation (cloud.google.com). This is a known trusted source and is used for reference only, not for direct code execution.

adk-engineer