The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides instructions for shaping the model's output style. These formatting constraints (e.g., removing preambles, leading with actions) do not attempt to bypass safety filters, jailbreak the model, or ignore system instructions. The skill explicitly notes that safety confirmation overrides brevity for destructive actions.
[DATA_EXFILTRATION]: The instructions do not involve reading sensitive files (such as credentials or environment variables) or making network calls to external destinations. No patterns for data exfiltration were found.
[REMOTE_CODE_EXECUTION]: There are no external package installations or remote script executions defined. The skill consists solely of natural language instructions for output formatting.
[OBFUSCATION]: All instructions are written in plain, human-readable text. There is no evidence of base64 encoding, hidden characters, homoglyph attacks, or other techniques used to conceal malicious intent.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize dynamic context injection syntax (e.g., shell command execution placeholders) to run commands at load time.

answer-first