The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation includes a dedicated file (SAFETY.md) that catalogs adversarial prompts and demonstrates the correct defensive behavior for the agent. While the file contains examples of injection attempts, they are presented as documentation for security evaluation purposes and are not intended to hijack agent behavior. The skill also acknowledges the risk of indirect injection from webhooks and mandates defensive normalization of external data.
[COMMAND_EXECUTION]: The skill facilitates file system operations and network communication for the BlueBubbles plugin. It includes explicit guardrails requiring path validation for downloads to prevent directory traversal and mandates the use of a specific URL builder helper to mitigate Server-Side Request Forgery (SSRF) risks.
[DATA_EXFILTRATION]: The skill manages sensitive configuration keys for the BlueBubbles server. It provides clear security guidelines instructing the agent to use proper configuration accessors and strictly prohibits the inclusion of credentials in logs, error messages, or outbound payloads.

bluebubbles-hardened