service-de-channel-consent-configure

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input for messaging prompts and keywords which are then interpolated into shell command templates.\n
  • Ingestion points: Stage 3 and 4 in SKILL.md collect user-supplied values for OptInPrompt, DoubleOptInPrompt, and various keyword fields like OptOutKeywords.\n
  • Boundary markers: The skill instructs the agent to wrap inputs in single quotes within double-quoted CLI arguments. While this prevents CLI parsing errors for spaces, it does not prevent the underlying shell from expanding command sequences like backticks or $() if they are present in the user input.\n
  • Capability inventory: The skill uses the sf CLI to query, create, and update records in the Salesforce org and writes results to temporary files in /tmp.\n
  • Sanitization: No explicit guidance is provided to the agent for sanitizing or escaping shell-sensitive characters from the user-provided prompt strings before command execution.\n- [COMMAND_EXECUTION]: The skill utilizes the sf CLI tool to perform its primary functions, which is the standard, expected method for managing Salesforce resources by the vendor.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 10:04 AM
Security Audit — agent-trust-hub — service-de-channel-consent-configure