The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with Salesforce environments by executing commands via the Salesforce CLI (sf). This includes querying data, managing org authentication, and performing record operations.
[EXTERNAL_DOWNLOADS]: The skill metadata recommends installing the @salesforce/cli package from the npm registry, a well-known and official service for this toolset.
[DYNAMIC_EXECUTION]: The skill supports executing Apex code (sf apex run) and making arbitrary REST API calls (sf api request rest). These high-privilege operations are primary intended features and are accompanied by explicit instructions requiring user confirmation before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Salesforce CRM records (via SOQL/SOSL) and local files. While this presents an ingestion surface for indirect prompt injection, the skill mitigates risk by instructing the agent to use structured JSON output and requiring human confirmation for any write or delete operations.
[DATA_EXPOSURE]: Commands such as sf org display --verbose can output sensitive authentication tokens. The skill provides clear instructions for the agent to never reveal access tokens, auth URLs, or refresh tokens in the chat response.

salesforce