The Agent Skills Directory

[Internal Configuration Access]: The skill reads and writes to specific directories (e.g., ~/.claude/plugins/config/claude-for-legal/) to manage legal matter logs and landscape data. This is standard behavior for maintaining context and state within its intended legal workflow environment.
[External Tool Integration]: It utilizes legal research platforms like Westlaw or CourtListener to verify procedural rules. The instructions include a 'No silent supplement' policy, which is a significant safety feature that prevents the agent from hallucinating legal information and instead prompts the user when tool results are insufficient.
[Indirect Prompt Injection Surface]: Because the skill analyzes external documents (subpoenas) provided by the user, it is technically exposed to data that could contain malicious instructions. However, the structured analysis workflow—which focuses on specific field extraction and predefined categories—minimizes the risk of the agent deviating from its task to follow embedded instructions.
[Professional Escalation and Guardrails]: The skill demonstrates security-conscious design by including explicit gates that stop processing and require human escalation for grand jury subpoenas or when the user role is identified as a non-lawyer.

subpoena-triage