The Agent Skills Directory

[SAFE]: No hardcoded credentials or secrets were found in the skill files.
Evidence: All JSON templates in the examples/connections/ directory use descriptive placeholders like <SNOWFLAKE_USERNAME> and <AZURE_APP_CLIENT_SECRET> to ensure users do not accidentally include real credentials in the skill definitions.
[COMMAND_EXECUTION]: The skill performs command-line operations using the Salesforce CLI and a local diagnostic script.
Evidence: Workflows involve sf data360 commands and node ~/.claude/skills/orchestrating-datacloud/scripts/diagnose-org.mjs for environment readiness checks. These are standard vendor-provided tools for Data Cloud management.
[PROMPT_INJECTION]: The system instructions are designed with clear task scoping and do not attempt to bypass agent guardrails.
Evidence: The SKILL.md provides explicit trigger conditions and delegation rules to other skills in the *-datacloud family without using adversarial patterns or bypass markers.

connecting-datacloud