audience-icp-filter

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted lead data (such as job titles and professional bios) which are then analyzed by the AI agent for semantic classification. This creates a surface for indirect prompt injection if the lead data contains adversarial instructions.
  • Ingestion points: Lead data provided in JSON format, specifically the jobTitle and shortBio fields processed in scripts/build.py and references/pass-two-review.md.
  • Boundary markers: The skill utilizes a structured JSON format and specific flags (--adjudicate) to separate input data from processing logic.
  • Capability inventory: The system performs regex-based pattern matching and semantic analysis by the AI model.
  • Sanitization: Input data is handled as strings for regex and NLP processing; no explicit sanitization for LLM-targeting injection patterns is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 06:54 AM
Security Audit — agent-trust-hub — audience-icp-filter