The Agent Skills Directory

[SAFE]: The skill is a robust toolkit for Salesforce data management that adheres to security best practices.
Data Integrity and Validation: The skill includes specialized Python scripts (soql_validator.py and validate_data_operation.py) designed to scan for security risks, such as hardcoded Salesforce IDs and PII patterns (SSNs, credit card numbers) before execution.
Command Execution: The skill wraps standard sf CLI commands for querying, creating, and deleting records. These are intended for use by a developer agent in a Salesforce environment and do not show signs of command injection or privilege escalation.
Data Isolation: Multiple documents and Apex templates (rollback-transaction.apex, cleanup-rollback-guide.md) promote the use of savepoints and cleanup scripts to ensure that test data does not persist in the target org, reducing the risk of data pollution or accidental exposure.
Author Trust: The skill is authored by Salesforce (forcedotcom), and the activities described align with their official documentation and developer tools.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill interacts with sensitive Salesforce org data, it includes proactive checks to prevent the use of PII in test datasets. No unauthorized network calls or remote exfiltration patterns were identified.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a standard attack surface for indirect prompt injection as it ingests CSV and JSON data from the file system to perform bulk operations. However, this is the primary purpose of the skill, and the risk is mitigated by syntax validation and the use of structured data templates.

platform-data-manage