The Agent Skills Directory

Indirect Prompt Injection: The skill processes external artifacts such as Privacy Impact Assessments (PIAs) and Data Processing Agreements (DPAs). Because these files may originate from various sources, they represent a surface where instructions could be embedded to influence the analysis. 1. Ingestion points: Files within the configured 'outputs' folder and the CLAUDE.md configuration file. 2. Boundary markers: The skill does not explicitly define delimiters for untrusted content. 3. Capability inventory: The skill can read local files and write to a specific configuration file. It does not possess network or code execution capabilities. 4. Sanitization: No specific sanitization or filtering of the ingested legal documents is described.
Local Configuration Management: The skill maintains state by reading from and writing to a local configuration file (~/.claude/plugins/config/claude-for-legal/privacy-legal/CLAUDE.md). This file contains sensitive data such as policy commitments and regulatory footprints. While necessary for operation, this interaction should be monitored to ensure only intended fields are modified.

policy-monitor