The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents a potential surface for indirect prompt injection through user-controlled inputs.
Ingestion points: User-provided sandbox names and IDs (07E prefix) captured during the interactive help menu or direct request flows.
Boundary markers: The skill lacks explicit delimiters or instructions for escaping user variables when they are interpolated into shell commands or SOQL strings.
Capability inventory: The skill is capable of executing shell commands via the Salesforce CLI, performing network operations through the Connect REST API, and executing SOQL queries via the Tooling API.
Sanitization: Although the skill instructions mention expected formats (e.g., alphanumeric, 10-character limit), there are no explicit requirements for the agent to sanitize or validate these inputs before they are included in executable commands.
[COMMAND_EXECUTION]: The skill executes administrative commands using the sf CLI. While these operations are consistent with the skill's stated purpose of sandbox management, the construction of these commands from external input represents a standard security surface that requires the agent to adhere strictly to the provided input constraints.

platform-sandbox-configure