The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using the Salesforce CLI (sf) and jq. To prevent command injection, the script validates the flow_developer_name and org_alias parameters against strict regular expressions (^[A-Za-z][A-Za-z0-9_]{0,79}$) before they are interpolated into command strings.
[INDIRECT_PROMPT_INJECTION]: The skill ingests metadata from the target Salesforce organization, such as ServiceChannel labels, to perform token resolution in Flow XML files. This surface for indirect injection is mitigated by the use of xml_escape and sed_repl_escape functions, which ensure that external data cannot corrupt the XML structure or hijack the sed templating process.
[SAFE]: The skill implements a production safety guard in scripts/deploy-and-report.sh that queries the Organization object. It verifies that the environment is a sandbox, trial, or developer edition, blocking deployment to production customer orgs by default.

service-omni-routing-flow-deploy