account-tier-scoring
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs legitimate business logic for account prioritization. It does not contain any evidence of credential theft, remote code execution, or unauthorized data exfiltration.
- [SAFE]: The skill processes external data signals (website visits, chat logs, social engagement) which constitutes a surface for indirect prompt injection.
- Ingestion points: CRM record data, website identification signals, and chat/support interaction logs (SKILL.md, references/intent-scoring-reference.md).
- Boundary markers: The skill uses a structured methodology with specific numeric thresholds and logic gates to parse data, though it lacks explicit "ignore embedded instructions" markers for the raw text of signals.
- Capability inventory: The agent is instructed to update CRM lead scores (SKILL.md Step 4) and post alerts to internal notification channels (SKILL.md Step 7).
- Sanitization: The skill includes a "Signal source verification" step (intent-scoring-reference.md Step 0.5) that validates email domains and assigns confidence levels to signals, acting as a filter against unverified or spoofed inputs.
Audit Metadata