The Agent Skills Directory

[SAFE]: The skill functions as a documentation and template resource for structuring AI projects. It does not contain malicious code, obfuscation, or unauthorized remote execution patterns.
[PROMPT_INJECTION]: The architecture utilizes local files in a data/ directory as a persistent memory layer. This creates a surface for indirect prompt injection where the agent might process instructions embedded in stored data (e.g., logs, project context, or inbox files). The skill assumes these files are managed within a trusted local environment.
[CREDENTIALS_UNSAFE]: The skill includes explicit security guidance in its 'Anti-Patterns' section, warning users against hardcoding API keys in agent files and recommending the use of environment variables as a safe alternative.

agentic-os