The Agent Skills Directory

[SAFE]: The document serves as a meta-skill, providing architectural guidelines, templates, and best practices for creating other skills. It does not contain executable code, remote downloads, or credential exposure.
[COMMAND_EXECUTION]: The skill mentions the use of local command-line tools such as wc -w for token counting and a local script render-graphs.js for visualizing DOT diagrams as SVGs. These are standard utility operations within the described development workflow and do not involve untrusted remote sources.
[PROMPT_INJECTION]: The skill uses strong instructional language (e.g., 'The Iron Law', 'REQUIRED BACKGROUND') and 'No exceptions' clauses to enforce a specific methodology (TDD). These are internal process constraints designed to ensure the agent follows a disciplined workflow and do not attempt to bypass safety filters, extract system prompts, or override the agent's core safety instructions.
[DATA_EXPOSURE]: References a local directory structure (~/.gemini/antigravity/skills/) for storing personal skills. This is a standard configuration path for the targeted agent system and does not constitute a leak of sensitive user information.

writing-skills