customer-win-back-sequencer

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting external, untrusted data to inform its decision-making and content generation.
  • Ingestion points: Data is collected from external sources via web_search and fetch_webpage tools during the research phases (Phase 1A, 1B, and 1C in SKILL.md).
  • Boundary markers: The skill instructions do not utilize boundary markers or explicit directives to ignore instructions that might be embedded within the retrieved web content.
  • Capability inventory: The skill generates personalized email sequences and writes detailed reports to the local file system (clients/ directory), which could be manipulated by injected content.
  • Sanitization: There is no evidence of validation, filtering, or escaping of the fetched external content before it is processed by the model for scoring or sequence generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 06:03 PM