legal-service-proposal
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides structured instructions for legal service drafting and pricing, referencing local workspace files such as 'law-practice-cn/CLAUDE.md' and 'references/pricing-proposal-framework.md' rather than untrusted remote sources.
- [SAFE]: The instructions include a specific data protection mandate that prohibits sending sensitive client information, such as budgets and historical rates, to external tools or MCPs.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data regarding client needs and matter details. 1. Ingestion points: Data enters the agent context through the argument hint ('[客户需求、事项类型、服务范围与团队约束]'). 2. Boundary markers: No specific delimiters are defined in the instructions to isolate user-provided data. 3. Capability inventory: The skill is limited to text generation and local file reading, with no defined capabilities for network-write operations or arbitrary command execution. 4. Sanitization: The instructions advise the agent to verify market data at runtime but do not specify technical sanitization methods.
Audit Metadata