gtm-stakeholder-mapping

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from CRM records, including activity logs and call transcripts, which could contain hidden malicious instructions.\n
  • Ingestion points: CRM contact inventories, activity logs, transcripts, and email patterns (SKILL.md Step 2, 3, 6).\n
  • Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the ingested CRM content.\n
  • Capability inventory: Artifact generation (Markdown, HTML, PDF), role classification, and deal strategy generation.\n
  • Sanitization: No sanitization or filtering methods for external CRM text inputs are described.\n- [DATA_EXFILTRATION]: The skill requires access to sensitive business data stored in CRM environments.\n
  • Sensitive data access: The skill requests broad access to contact records, titles, engagement levels, and opportunity details (SKILL.md Input Contract).\n
  • Network access: The skill identifies public sources as an intelligence source alongside CRM data, implying network operations occur while sensitive data is in the agent context.\n- [DYNAMIC_EXECUTION]: The skill includes instructions for generating PDF artifacts using Python code at runtime.\n
  • The artifact generation section (SKILL.md) specifies the use of the 'reportlab' library to dynamically build reports based on analyzed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 04:19 PM
Security Audit — agent-trust-hub — gtm-stakeholder-mapping