The Agent Skills Directory

[SAFE]: The skill acts as an instructional framework for developers. It explicitly advocates for secure practices, such as using environment variables for sensitive credentials and implementing user confirmation steps before performing actions.
[DYNAMIC_CONTEXT_INJECTION]: The reference.md file documents the !command syntax for preprocessing context. The examples provided (e.g., git status, gh pr diff) are standard development tools used to illustrate legitimate functionality without evidence of malicious use.
[INDIRECT_PROMPT_INJECTION]: The skill uses an interactive "Discovery Interview" to collect requirements from the user. While this process ingests external input to generate instructions, it includes built-in safeguards, such as a confirmation round where the agent summarizes the proposed workflow for human review before any files are created.
[EXTERNAL_DOWNLOADS]: The reference.md file contains links to official documentation for Claude Code hosted on the trusted domain code.claude.com.

skill-builder