The Agent Skills Directory

[SAFE]: The skill instructions and referenced guardrails emphasize safety, user approval, and refusal of insecure recommendations. No indicators of prompt injection, data exfiltration, or malicious execution were found.
[PROMPT_INJECTION]: The content contains defensive instructions that reinforce security boundaries rather than attempting to bypass them. It explicitly directs the agent to refuse changes that would weaken security controls.
[DATA_EXPOSURE]: The skill does not access sensitive system paths or hardcoded secrets, and it contains no network communication capabilities.
[REMOTE_CODE_EXECUTION]: No external dependencies are fetched, and there are no commands for remote script execution or dynamic code evaluation.

duck-teach