The Agent Skills Directory

[DATA_EXFILTRATION] (SAFE): The package_skill.sh script includes a robust exclusion list (EXCLUDE_PATTERNS) that prevents the accidental inclusion of sensitive files like .env, .pem, id_rsa, and credentials in distributed zip files.
[COMMAND_EXECUTION] (SAFE): All scripts (init_skill.sh, validate_skill.sh, package_skill.sh) utilize standard system utilities such as awk, sed, zip, and find for local file operations. The scripts perform directory existence checks and regex-based name validation to prevent path traversal or unintended file creation.
[PROMPT_INJECTION] (SAFE): The content is purely instructional and metadata-focused. It contains no instructions directed at the agent to ignore safety protocols or override its core identity.
[EXTERNAL_DOWNLOADS] (SAFE): The skill does not fetch external code or dependencies at runtime; it operates entirely with local templates and system commands.
[OBFUSCATION] (SAFE): The code and documentation are transparent and well-commented, with no evidence of encoding or hidden characters intended to deceive analysis.

skill-creator