amendment-history

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill analyzes external contract files, which are ingestion points for potentially untrusted data. While it lacks explicit boundary markers to isolate document text from the agent's logic, it uses section-specific indexing and quoting to structure its output. The agent's capabilities include reading and writing to matter-specific directories to manage the analysis workflow.\n- Matter Workspace Management: The skill interacts with local configuration paths (at ~/.claude/) to maintain matter context and store work product. This ensures that legal analysis is properly categorized by matter and stored in designated locations, facilitating organization and privilege management.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 07:01 PM
Security Audit — agent-trust-hub — amendment-history