The Agent Skills Directory

[SAFE]: The skill performs all Salesforce operations through a dedicated Model Context Protocol (MCP) server, which abstracts the authentication and raw network communication, maintaining a secure boundary between the agent and the target organization.
[SAFE]: It features a comprehensive validation framework using bundled Python scripts (soql_validator.py, mcp_validator.py, etc.) that perform pre-flight checks for PII, SOQL syntax errors, and potential performance issues before any data is sent to the org.
[SAFE]: A PreToolUse hook (pre-mcp-validate.py) is implemented to intercept and validate tool calls, providing an automated layer of security that warns the user if an operation is risky or malformed.
[SAFE]: The skill explicitly instructs the agent to follow Salesforce security best practices, such as using WITH USER_MODE and WITH SECURITY_ENFORCED to respect field-level security and sharing rules.
[SAFE]: All provided assets, including Apex factories, CSV templates, and JSON tree examples, contain only generic test data and placeholder identifiers, with no evidence of credential exposure or sensitive information leakage.

sf-data