service-concierge-portal-generate

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the sf (Salesforce) CLI and standard system tools (curl, python3, jq) to automate a multi-step deployment process. These commands are used for querying org data, activating agents, and deploying metadata bundles. Subprocess calls to python3 -c are employed safely for JSON parsing and file modifications.
  • [INDIRECT_PROMPT_INJECTION]: The skill accepts user inputs such as the site name and branding colors. The site name is explicitly sanitized to alphanumeric characters only before being interpolated into CLI commands and API payloads, mitigating potential command injection risks. The capability inventory is restricted to Salesforce configuration tasks.
  • [SAFE]: All network operations and external API requests target official Salesforce infrastructure (e.g., *.salesforce.com, *.my.site.com, *.salesforce-scrt.com). The runbook includes clear disclosures regarding security setting changes (like CSP relaxation) that are necessary for the Concierge feature's functionality, and it includes manual checkpoints for user review of critical org permissions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 04:35 PM
Security Audit — agent-trust-hub — service-concierge-portal-generate