The Agent Skills Directory

[SAFE]: The skill functions as a security education and utility toolkit, providing defensive patterns and implementation templates without malicious logic.- [INDIRECT_PROMPT_INJECTION]: The skill has ingestion surfaces through tools like WebFetch and Read. However, it explicitly addresses these risks by providing robust mitigation patterns including context separation (detailed in references/context-separation.md) and prompt auditing (detailed in scripts/prompt_builder.py) to prevent accidental execution of embedded instructions.- [COMMAND_EXECUTION]: The skill includes a vulnerability scanning utility (scripts/scan-vulnerabilities.sh) that executes standard development tools such as npm audit, pip-audit, and bandit. This behavior is consistent with the skill's stated purpose of security auditing and follows best practices for dependency management.- [EXTERNAL_DOWNLOADS]: References to external libraries and resources (e.g., Microsoft Presidio, Zod, Pydantic) are informational and intended for developer setup, targeting well-known and trusted software repositories.

security-patterns