renewal-radar

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external communication channels to generate risk assessments, creating a potential surface for indirect prompt injection.
  • Ingestion points: Content from email threads, internal Slack messages, and uploaded documents are used to identify risk signals (e.g., support escalations, sentiment).
  • Boundary markers: The skill instructs the agent to ground custom field names before use and to cite all values exactly as they were queried from the source.
  • Capability inventory: Includes reading Salesforce data via dispatch_readonly, searching communication tools, and updating Salesforce records via dispatch (specifically for 'Hygiene' tasks like updating dates or next steps).
  • Sanitization: The skill relies on structured JSON templates for the display_widget tool to present data, which limits the potential for raw data to influence agent behavior outside of the defined scoring logic.
  • [DATA_EXFILTRATION]: The skill accesses sensitive business data (opportunities, emails, documents). However, this data remains within the user's session and is only output via authorized platform mechanisms such as the display_widget or markdown text. There are no network requests to non-whitelisted or third-party domains.
  • [COMMAND_EXECUTION]: The skill uses the dispatch tool for updating Salesforce opportunities in the 'Hygiene' section. The instructions explicitly require the agent to present proposed changes for user confirmation before executing any write operations, serving as a critical safety control.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 06:39 PM