The Agent Skills Directory

Workspace Data Management: The skill performs file creation, modification, and directory movement (archiving) within the ~/.claude/ configuration directory to maintain separate workspaces. These operations are restricted to the local environment and are necessary for the skill's stated purpose of workspace management.
Data Isolation Controls: The instructions include explicit logic for "Cross-matter context" flags to prevent data leakage between different workspaces. This is a positive security design pattern for skills handling multi-client or multi-project environments.
Input Validation: The skill instructs the agent to enforce specific formatting for workspace identifiers (lowercase slugs with hyphens) and requires the agent to show changes and seek user confirmation before writing to the file system.
Indirect Prompt Injection Surface: The skill collects matter-specific metadata (client names, counterparty details, matter types) through an intake process and stores it in workspace configuration files. While this creates a surface where external data is stored in the agent's context, the skill does not execute this data.
Ingestion points: User-provided input during the new <slug> intake interview described in SKILL.md.
Boundary markers: Explicit instruction to "Show the user what changed and confirm before writing."
Capability inventory: File write (matter.md, history.md, notes.md) and directory move (archiving) within the ~/.claude/plugins/config/claude-for-legal/ path.
Sanitization: Instruction to validate slugs as "lowercase with hyphens" to prevent improper path construction.

matter-workspace