The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it parses organizational structures from external sources such as PDF documents, website URLs, and CSV files in Step 1. This untrusted data is then used to generate Salesforce records and hierarchy visualizations.
Ingestion points: Step 1 involves reading content from user-provided PDF paths, website URLs, and CSV files.
Boundary markers: The skill mitigates risks by requiring explicit user confirmation of the parsed structure in Step 3 ("Confirm desired structure with user") and a pre-write preview in Step 6 before any records are modified.
Capability inventory: The skill has permissions to perform Salesforce REST API operations (POST, PATCH, DELETE) for Account and BusinessProfile objects, Tooling API writes for RecordType creation, and local file-system writes for the completion report in Step 10.
Sanitization: The skill performs noise filtering on PDF content (e.g., page numbers, copyright symbols) as described in the parsing rules, though it relies primarily on human review to validate the logical structure before execution.
[COMMAND_EXECUTION]: The skill utilizes the sf CLI and Bash for administrative automation. These actions are strictly scoped to the primary purpose of the skill—Salesforce environment configuration—and are performed through established platform tools.
[PRIVILEGE_ESCALATION]: The skill provides instructions for creating Account RecordTypes via the Salesforce Tooling API in the event of missing prerequisites. While administrative in nature, this behavior is a documented part of the institutional hierarchy setup and requires explicit user consent before proceeding with writes.

education-cloud-multi-campus-configure