diligence-issue-extraction
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze a large volume of untrusted external documents from data rooms, which is a primary surface for indirect prompt injection attacks.
- Ingestion points: Processes files from external sources including data rooms, Feishu, and Jianguoyun via MCP tools, as well as local configuration files like
deal-context.md(SKILL.md). - Boundary markers: The skill defines strict extraction categories (Material Contracts, IP, Labor, etc.) and uses a mandatory internal memo format to structure the output.
- Capability inventory: Reads data room files, classifies legal issues, generates summary reports, and writes output to specific matter folders (e.g.,
~/.claude/plugins/config/.../matters/). It can also hand off tasks to other configured AI tools. - Sanitization: Implements a rigorous source attribution system (e.g.,
[yuandian检索],[用户提供]), uses specific warning tags for unverified information ([需核实]), and includes a 'Prohibition of Silent Supplementation' policy to prevent the model from filling data gaps with unverified knowledge.
Audit Metadata