The Agent Skills Directory

[SAFE]: The skill is a collection of high-quality engineering instructions designed to improve the accuracy and safety of AI-driven development. It implements 'Safety Gates' that require the agent to perform backups (e.g., using cp) and check for dependencies (e.g., using grep) before making system changes.
[PROMPT_INJECTION]: The instructions use persona-setting techniques to act as a 'professional engineer' but do not contain instructions to bypass safety filters, ignore system prompts, or reveal sensitive configuration data.
[DATA_EXFILTRATION]: While the skill mentions using tools like curl, Bash, and WebSearch, these are presented as legitimate means for debugging (e.g., checking API responses or searching for error messages). There are no hardcoded credentials or instructions to send sensitive data to unauthorized external domains.
[COMMAND_EXECUTION]: The skill encourages the use of command-line tools for verification (e.g., ls -la, cat, node -v). These are used within the context of the agent's existing toolset to fulfill the user's engineering requests and do not introduce arbitrary or hidden execution paths.

yes-ja