The Agent Skills Directory

[PROMPT_INJECTION]: The instructions establish a professional software development persona focusing on craftsmanship and ownership. The language is purely instructional and does not attempt to bypass safety filters or override system instructions.
[DATA_EXPOSURE]: The skill provides explicit guidance in references/domains/security.md to avoid hardcoding secrets and suggests using environment variables or secrets managers. It mentions sensitive paths like .ssh/id_rsa and .env only as examples of what to protect, following security best practices.
[COMMAND_EXECUTION]: Static analysis flagged destructive commands; however, these are contained within educational examples in references/domains/security.md to demonstrate vulnerabilities like command injection and insecure permissions (e.g., rm -rf / and chmod 777) so that the agent can recognize and avoid them. The skill does not instruct the agent to execute these commands maliciously.
[EXTERNAL_DOWNLOADS]: The skill does not contain any commands to fetch or execute remote code. It focuses on local development tasks and repository management.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or unsafe secret handling practices were found. The documentation actively encourages the use of secure secret management tools.

vibe-coding