diligence-issue-extraction
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill interacts with local files for configuration and context. Evidence: It reads diligence parameters and materiality thresholds from
~/.claude/plugins/config/claude-for-legal/corporate-legal/CLAUDE.mdand deal-specific context from~/.claude/plugins/config/claude-for-legal/corporate-legal/deals/[code]/deal-context.md. It writes findings to paths such as~/.claude/plugins/config/claude-for-legal/corporate-legal/matters/<matter-slug>/. While this is central to the skill's function, it exposes sensitive information within the user's configuration files to the AI agent. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents from external Virtual Data Rooms.
- Ingestion points: Documents retrieved via MCP tools for Box, Intralinks, or Datasite.
- Boundary markers: The instructions lack explicit delimiters or guidance to ignore malicious instructions embedded in the documents during extraction.
- Capability inventory: The skill possesses read/write access to the local file system and interacts with legal research tools.
- Sanitization: There is no evidence of sanitization or content validation for the data extracted from the documents.
Audit Metadata