LPM Core Plugin for Claude

Most discussions about AI in legal practice focus on what AI will replace. This project starts from a different premise: legal project management is the discipline best positioned to author AI skills, because skills architecture — trigger conditions, sequential workflows, validation gates, progressive disclosure — is project management methodology expressed as markdown.

This plugin is the implementation. Fourteen skills covering the full LPM matter lifecycle: initial intake and scoping, matter planning, budget and fee management, billing cycle management, scope control, status reporting, risk and issues management, stakeholder communications, resource planning, local counsel management, timeline generation (with full dependency logic and critical path), collaboration platform configuration, continuous improvement, and document approval tracking.

All fourteen skills are built and tested against realistic synthetic scenarios. Apache 2.0. The repo is at github.com/legalopsconsulting/lpm-skills. Install as a plugin or clone and install manually — the README has both methods.

Design principles across all fourteen skills: email-first input with structured output; dual mode operation (manual paste-and-process, and M365 MCP connected for monitoring Outlook and reading/writing SharePoint); methodology encoded, legal content belongs to the deploying firm; .docx as primary output because matter records belong in the matter folder, not the chat window.

The plugin is part of an expansion pack architecture. LPM Core installs independently; practice-area plugins (LPM for M&A is the first, with five skills for cross-border transactions and corporate reorganisations) add on top without rebuilding the core. v1 complete.

Tested against synthetic scenarios, not yet battle-tested on live matters. Feedback welcome.

• 14-skill architecture covering the full LPM matter lifecycle — intake through to close
• Email-first input: paste emails, call notes, or unstructured updates; receive structured, actionable output
• Dual mode: manual (paste-and-process, works on any Claude plan) and M365 MCP connected (monitors Outlook, reads/writes SharePoint — available on Team and Enterprise)
• Scope control with two-layer model: Standard for every matter, Extended for complex matters — automated capture surfaced for confirmation, not creation
• RAID log methodology with decision extraction directly from email correspondence
• RAG-rated status reporting with silence-as-signal detection — workstreams that haven't reported get flagged, not ignored
• Timeline generation with all four dependency types (FS, FF, SS, SF), lag/lead values, critical path, near-critical path, and what-if cascade mode
• .docx output as default — matter records belong in the matter folder, not in the chat window
• Apache 2.0 — free to use, modify, and deploy
• Expansion pack architecture: LPM Core as base install; practice-area plugins (M&A first) add on top

Synthetic testing gets you so far. The skills need real matter input — messy emails, incomplete updates, the kind of material that doesn't match the expected format. If you're an LPM or legal ops professional with material to test against, I'd welcome the feedback. Critical is more useful than confirmatory.

If you're thinking about the build-vs-buy question for institutional methodology — firm-owned skills versus vendor playbooks — I'd like to hear your view.