customer-persona

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill exhibits no malicious behaviors such as network exfiltration, remote code execution, or obfuscation. It functions as a set of logical instructions for the agent to process business information.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided inputs to generate personas and updates a shared knowledge base file.
  • Ingestion points: The skill reads from knowledge/icp/personas.md and knowledge/markets/positioning.md to establish context, and processes user research data provided during the session.
  • Boundary markers: Employs structured Markdown templates and specific summary card formats to isolate generated persona components.
  • Capability inventory: Performs file write operations to output/customer-persona/ and potentially knowledge/icp/personas.md (metadata in SKILL.md).
  • Sanitization: Implements a strict validation layer requiring source tags (e.g., call recordings, tickets, interviews) for all data entries and explicitly forbids the agent from inventing customer quotes or objections. It also requires a manual review and user approval of diffs before any shared knowledge file is updated.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 06:28 AM
Security Audit — agent-trust-hub — customer-persona