marketing-claims-review

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFE
Full Analysis
  • File Access and Workspace Management: The skill reads and writes to specific local paths within the ~/.claude/plugins/config/claude-for-legal/ directory. This is used to load marketing standards and manage organized workspaces (matters), which is consistent with the skill's purpose as a professional legal automation tool.
  • Source Attribution and Verification Framework: The instructions include a detailed system for tagging legal citations (e.g., [settled], [verify], [verify-pinpoint]). This encourages accuracy and alerts the user to citations that require manual verification against primary legal sources.
  • Structured Processing of External Data: While the skill ingests potentially untrusted marketing copy, it mitigates indirect instruction risks through a strict, multi-step review workflow. By forcing the agent to first extract claims and then classify them against a pre-defined taxonomy, it limits the likelihood of embedded text overriding the agent's core logic.
  • Professional Guardrails: The skill includes an explicit safety gate for non-lawyer users, requiring specific disclaimers and summaries before any copy can be approved. This promotes responsible use in professional environments where legal advice is required.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 12:40 AM