skills/swan-gtm/gtm-skills/brief/Gen Agent Trust Hub

brief

Pass

Audited by Gen Agent Trust Hub on Jul 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from HubSpot records, meeting notes, and call recordings to synthesize a brief. This creates a surface for indirect prompt injection, where malicious instructions hidden in the processed data could influence the agent's behavior.
  • Ingestion points: Data retrieved through hubspot-get-engagements, hubspot-search-objects, and contact records.
  • Boundary markers: The instructions do not define explicit delimiters or headers to separate external data from system instructions.
  • Capability inventory: The agent has access to swan-execute-code and CRM search tools.
  • Sanitization: No specific sanitization or filtering logic is prescribed for the content retrieved from external meeting notes or call records.
  • [COMMAND_EXECUTION]: The skill uses swan-execute-code to process JSON files containing HubSpot engagement history. This pattern involves executing code at runtime to parse, filter, and extract specific notes from tool outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 28, 2026, 04:39 PM
Security Audit — agent-trust-hub — brief