The Agent Skills Directory

[SAFE]: The skill provides a professional playbook for building Salesforce agents. It includes significant guidance on implementing Field-Level Security (FLS) and granular permissions to ensure agents follow the principle of least privilege, as detailed in the 'Permissions & FLS' section of BEST-PRACTICES.md.
[COMMAND_EXECUTION]: The skill uses standard Salesforce CLI (sf) commands for metadata deployment and running Apex scripts, as described in SKILL.md. These are standard tools for Salesforce development and are executed within the developer's authenticated environment.
[EXTERNAL_DOWNLOADS]: The skill fetches demo implementations and configuration from internal Salesforce Git repositories (git.soma.salesforce.com) and installs the official Salesforce CLI via NPM, as noted in SKILL.md. These resources originate from the vendor's own infrastructure and well-known services.
[INDIRECT_PROMPT_INJECTION]: The agents constructed using these templates are designed to ingest untrusted data from Salesforce Case and MessagingSession records. The skill provides architectural guidance to mitigate risks, such as enforcing Human-in-the-Loop (HiL) confirmation buttons for any write operations or transactions, as documented in CLT-GUIDE.md.
Ingestion points: Salesforce Case and MessagingSession records, including customer messages.
Boundary markers: The skill suggests using sort orders and prerequisite language in BEST-PRACTICES.md to control agent behavior and sequence execution.
Capability inventory: The agents use Apex @InvocableMethod actions and Salesforce Flows to interact with CRM data, as defined in CLT-GUIDE.md and SUBAGENT-TEMPLATE.md.
Sanitization: The skill relies on standard Salesforce platform security and LLM reasoning to process user-supplied data safely.

build-agentforce-service-demo