The Agent Skills Directory

[SAFE]: The skill provides guidelines and tools for creating other AI agent skills. The SKILL.md and references/ files promote security best practices, including a mandatory security policy block for skills intended for the Skillmark benchmark, which covers prevention of prompt injection, jailbreaks, and data exfiltration.
[SAFE]: Secret management guidance in references/script-quality-criteria.md and references/validation-checklist.md follows industry standards by instructing developers to use .env files and hierarchical environment variables, which is a recognized safe practice for handling credentials.
[SAFE]: The utility scripts (scripts/init_skill.py, scripts/package_skill.py, scripts/quick_validate.py) perform legitimate local development tasks such as templating, directory initialization, and validation of skill metadata. They do not execute remote code or perform suspicious file system operations.
[SAFE]: Documentation regarding plugin marketplaces (references/plugin-marketplace-*) describes official platform features for distributing extensions, including support for pinned versions via commit SHAs and standard authentication tokens for private repository access.
[SAFE]: No obfuscation, prompt injection attempts, persistence mechanisms, or privilege escalation patterns were found in any of the 28 analyzed files.

skill-creator