objection-pattern-detector

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data ('lost deal notes' and 'won deals') which presents a surface for indirect prompt injection if the processed documents contain malicious instructions.
  • Ingestion points: Processes deal notes and sales data provided at runtime as described in the SKILL.md instructions.
  • Boundary markers: The instructions lack specific delimiters or instructions for the agent to ignore potentially malicious content embedded within the deal notes.
  • Capability inventory: The skill utilizes standard text analysis capabilities to identify patterns and generate frameworks based on the ingested content.
  • Sanitization: No explicit sanitization or validation protocols are defined for the incoming deal note data.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 07:25 AM
Security Audit — agent-trust-hub — objection-pattern-detector