The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the sf (Salesforce) CLI to execute data queries and create records within the target Salesforce organization. This is the primary intended function of the skill for managing Omni-Channel settings.
[SAFE]: The script scripts/deploy-and-report.sh implements a robust production guardrail. It queries the target organization's metadata (IsSandbox, TrialExpirationDate, OrganizationType) and blocks execution if the target is a production instance, preventing accidental configuration changes to live environments.
[SAFE]: Environment variables SUPERVISOR_ACTIONS and SUPERVISOR_TABS are used to configure the deployment, but the script validates every requested type against a hardcoded whitelist (SAFE_ACTIONS and SAFE_TABS). This prevents the injection of arbitrary or malicious values into the CLI commands.
[COMMAND_EXECUTION]: The test utility scripts/tests/_bootstrap.py uses subprocess.run() to execute the main deployment script. This is performed within a controlled test environment to validate the skill's logic and does not pose a risk during normal operation.
[INDIRECT_PROMPT_INJECTION]: The skill processes data returned from the Salesforce API. The risk of indirect injection is mitigated by the following factors:
Ingestion points: Data is ingested via sf data query results in scripts/deploy-and-report.sh.
Boundary markers: The skill processes data as structured JSON using jq, ensuring clear boundaries between data and logic.
Capability inventory: The skill can create records in the Salesforce org but lacks general shell or network access.
Sanitization: Inputs are strictly validated against a hardcoded list of allowed configuration types before being used in any write operations.

service-omni-supervisor-surface-deploy