termination-review
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to process user-supplied termination scenarios and documentation. This creates a potential surface where embedded instructions in those external inputs could attempt to influence the agent's analysis. This risk is managed by the skill's structured multi-step checklist and the requirement for professional legal oversight.
- Ingestion points: User-provided text descriptions and attached documentation (Step 1).
- Boundary markers: The skill employs a strict five-step workflow and explicit checklist categories to maintain structured processing.
- Capability inventory: The skill utilizes file system writing for matter management and invokes external legal research tools (Step 3).
- Sanitization: The skill focuses on mapping inputs to specific legal flags rather than executing instructions contained within the user-provided content.
- Application-Scoped File Access: The skill reads and writes to the
~/.claude/plugins/config/directory. These operations are localized to the application's configuration and matter-tracking environment, which is standard for maintaining state within a specialized agent workflow.
Audit Metadata