log-activity

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains instructions that direct the agent to override or disregard specific environment context regarding its authorization status. It mandates that the agent must ignore status readouts claiming "not connected" or "needs re-auth" and instead prioritize attempting a direct tool call. This is an operational override intended to improve reliability by bypassing potentially stale metadata about the connector's state.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of user-provided transcripts or summaries and interpolates this content into Salesforce Task descriptions. This creates a surface where instructions hidden within a transcript could attempt to influence the agent's drafting or execution logic.
  • Ingestion points: Untrusted user input provided as meeting summaries or call transcripts (SKILL.md).
  • Boundary markers: The skill implements a critical human-in-the-loop checkpoint, requiring explicit user confirmation ("Ready to log this to Salesforce?") before the dispatch tool is used to write data.
  • Capability inventory: The skill possesses the capability to write to the Salesforce Task object using the dispatch tool (POST request to /sobjects/Task).
  • Sanitization: No specific technical sanitization or escaping of the transcript content is defined; the skill relies on the confirmation prompt to mitigate malicious content in the generated draft.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 06:39 PM