The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses natural instructional language to guide the agent toward higher quality output. It does not contain any attempts to bypass safety filters, override system instructions, or extract sensitive prompt data.
[DATA_EXFILTRATION]: No network operations or sensitive file access patterns were detected. The skill contains educational examples of secure vs. insecure coding (e.g., login endpoints, JWT handling), but these are provided as markdown documentation for the agent to follow, not as executable commands for exfiltration.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute any remote scripts. All code examples provided are within markdown blocks for instructional purposes only.
[COMMAND_EXECUTION]: The skill mentions CLI tools like grep and git as examples of how a developer might study a codebase, but it does not execute them automatically or in an unsafe manner.
[OBFUSCATION]: The skill is written in clear, plain-text markdown. There are no signs of encoding, hidden characters, or homoglyph-based attacks.
[SAFE]: The skill serves as a 'craftsman mode' for the agent, encouraging thorough design, security checklists (e.g., XSS/SQL injection prevention), and simple, maintainable architecture. It represents standard developer tooling guidelines.

deep-think