The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Bash to invoke the Salesforce CLI (sf) and a local Node.js script. These operations are limited to authenticated API requests and internal data classification, adhering to the principle of least privilege by performing only read operations during validation.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Salesforce API endpoints. A mandatory evidence chain analysis shows:
Ingestion points: Data enters through sf api request rest which fetches feature toggle statuses from Salesforce.
Boundary markers: API responses are captured to temporary JSON files and processed via a standalone script rather than being directly interpolated into the primary prompt.
Capability inventory: The skill uses Bash for command execution and AskUserQuestion for user interaction.
Sanitization: The classify-readiness.mjs script performs strict JSON parsing and deterministic mapping of specific feature flags, preventing the agent from being influenced by unexpected or malicious content within the API response.
[DATA_EXFILTRATION]: The skill explicitly instructs against the extraction of accessToken from the CLI, ensuring that session credentials are not leaked into shell history or environment variables. All network communication is proxied through the authenticated Salesforce CLI, targeting only the user's own Salesforce organization.

service-itsm-agentic-setup-agentforce-studio-validate