The Agent Skills Directory

[SAFE]: No security threats detected. The skill acts as an educational toolkit and template library for Salesforce developers.
[PROMPT_INJECTION]: The skill provides meta-instructions and guidelines for building safe and robust AI agents. This includes teaching developers how to implement AI disclosure, prevent system prompt leakage, and maintain conversational boundaries. These instructions are best-practice examples for the development process and do not attempt to override the host agent's safety protocols.
[DATA_EXFILTRATION]: The skill uses the Salesforce CLI (sf) and curl to interact with the user's Salesforce instance. All network operations documented are part of the standard Salesforce developer workflow, such as provisioning Data Libraries or deploying metadata to the user's authenticated environment. No sensitive information is transmitted to third-party or untrusted domains.
[COMMAND_EXECUTION]: The skill provides templates for Apex code and Salesforce CLI commands. These are intended to be reviewed, modified, and executed by the developer within their own secure environment. There is no evidence of arbitrary or silent command execution.

developing-agentforce