objection-pattern-detector
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- Indirect Prompt Injection (LOW): The skill is designed to ingest and analyze external deal notes, which creates a surface for indirect prompt injection if those notes contain malicious instructions. Evidence Chain: 1. Ingestion points: External 'lost deal notes' and 'won deals' data mentioned in SKILL.md. 2. Boundary markers: None present; there are no instructions to the agent to ignore embedded commands within the deal notes. 3. Capability inventory: No active capabilities such as file-writing, network access, or command execution are defined in the skill files. 4. Sanitization: No sanitization or escaping mechanisms are described for the input data.
Audit Metadata