The Agent Skills Directory

[SAFE]: The skill contains technical documentation and code examples for a development framework. It focuses on best practices for Domain-Driven Design, Hexagonal Architecture, and security.
[DYNAMIC_EXECUTION]: Reference material in references/sandbox-execution.md and references/inference.md describes how the framework handles code execution. Specifically, it provides a SandboxSpec for governed out-of-process execution (including containerized isolation for untrusted code) and warns about the risks of using pickle for model deserialization. These are documented capabilities and security warnings rather than malicious behaviors.
[CREDENTIALS_UNSAFE]: The documentation references how to manage credentials using SecretsPort, Vault, and SecretStr in Pydantic models. It explicitly advises against hard-coding secrets in application code, promoting secure secret management practices.
[INDIRECT_PROMPT_INJECTION]: The framework is designed to process external data through various ports (database, storage, inference). While this creates an attack surface for indirect prompt injection in applications built with the framework, the documentation includes guidance on sanitization and scrubbing.

forze-skills