policy-drafting
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Local Configuration and Data Access: The skill is designed to read from and write to specific paths within the
~/.claude/plugins/config/directory, such asCLAUDE.mdand matter-specific folders. This mechanism allows the agent to customize policy drafts based on a company's unique jurisdictional footprint and active legal matters. The file operations are confined to this local plugin environment and are necessary for the skill's stated purpose of providing contextually relevant legal drafting. - Processing of Untrusted Contextual Data: As part of its workflow, the skill ingests content from external markdown files to determine drafting parameters. This creates a surface for indirect prompt injection, where instructions embedded within those source files could potentially influence the agent's behavior or output style. While this warrants consideration during the preparation of configuration files, the skill includes a strong procedural requirement for review by a licensed legal professional before any draft is finalized or implemented, effectively managing the risks associated with automated content generation.
Audit Metadata