The Agent Skills Directory

[SAFE]: All generated outputs include explicit 'DRAFT FOR INTERNAL LEGAL REVIEW' headers and mandatory '[review]' flags for all key decision points, ensuring that the AI output is treated as a working draft rather than a finished policy.
[SAFE]: File system interactions are limited to the skill's own plugin configuration directory (~/.claude/plugins/config/claude-for-legal/) and matter-specific workspaces, which is standard behavior for managing user-specific legal matter data.
[SAFE]: External interactions are restricted to web searches for authoritative legal guidance (e.g., ABA, state bar policies), and the skill provides a 'Sources' block to document every external reference used in the draft.

policy-starter