The Agent Skills Directory

Legal Ethics and Compliance Integration: The skill includes explicit instructions to adhere to jurisdiction-specific rules, such as PD 57AC for England & Wales, to prevent improper AI-assisted drafting of witness statements. It also references MR 3.1 regarding candor about weak legal arguments, demonstrating a high degree of alignment with professional legal standards.
Strict Citation Controls: The implementation of an exhaustive citation extraction and verification workflow, including markers like [VERIFY] and [CITE NEEDED], is designed to mitigate the risks of hallucinated or misgrounded citations in legal work product.
Controlled Context Access: The skill accesses specific local configuration files located at ~/.claude/plugins/config/claude-for-legal/ to maintain consistency with firm style and case theory. These paths are scoped to the agent's internal plugin structure rather than sensitive system directories.
Conflicts and Intake Validation: A mandatory gate requires the skill to verify a matter's presence in a local _log.yaml before proceeding, ensuring that conflicts checks and intake procedures are followed before substantive work is performed.
Source Attribution and Tooling: The skill mandates the tagging of all citations by source (e.g., specific legal research tools versus model knowledge) and provides clear guidance to non-lawyers regarding the legal consequences of filing documents without professional review.

brief-section-drafter