Get Access
Prompt Multi-domain Claude · GPT-4 · MCP

Best Prompts & MCP of the Week — May 11, 2026

This week: a contract risk scanner for legal teams and the Filesystem MCP for secure document workflows without IT tickets.

Prompt of the week

The Contract Risk Scanner: Rapid Red-Flag Review for Non-Lawyers

What it does: Analyzes any commercial contract and surfaces liability exposure, missing standard clauses, and negotiation leverage points in plain language. Built for in-house legal, procurement, and finance teams who need a first-pass review before escalating to outside counsel.

When to use it:

  • Reviewing vendor MSAs or SaaS agreements before signature
  • Flagging risk in NDAs received from counterparties during M&A due diligence
  • Preparing a summary for a business stakeholder who needs to understand what they are actually signing

The prompt:

You are a senior commercial contracts attorney with 15 years of experience advising mid-market and enterprise companies. Review the contract text below and produce a structured risk report.

**Contract text:**
[PASTE FULL CONTRACT TEXT HERE]

**Company context:**
- Our company name: [COMPANY NAME]
- Our role in this agreement: [BUYER / SELLER / LICENSOR / LICENSEE / OTHER]
- Industry: [INDUSTRY]
- Contract value or ARR: [VALUE OR "UNKNOWN"]

**Deliver the following sections:**

1. **Executive Summary** (3–5 sentences): What is this agreement, what does it obligate us to do, and what is the headline risk level (Low / Medium / High)?

2. **Red Flags** (bulleted list): Clauses that are unusual, one-sided, or missing entirely. For each, cite the section number, explain the risk in plain English, and suggest alternative language.

3. **Indemnification & Liability Exposure**: Summarize caps, carve-outs, and uncapped obligations. State whether our exposure is bounded.

4. **Termination & Renewal Terms**: Notice periods, auto-renewal traps, and termination-for-convenience rights.

5. **Data & IP Ownership**: Who owns work product, data, and derivative works?

6. **Recommended Next Steps**: Three prioritized actions before we sign.

Use plain English throughout. Do not give legal advice; flag where outside counsel review is essential.

Tips:

  • Run this on GPT-4o or Claude 3.7 with a large context window — paste the full contract, not excerpts, to catch cross-reference risks between clauses.
  • Add a line specifying your governing law jurisdiction (e.g., “Governing law: New York”) to get more precise clause benchmarking.
  • For NDAs specifically, add “Flag any non-standard definition of Confidential Information and any survival periods exceeding 3 years” to the Red Flags instruction.

MCP deployment of the week

Filesystem MCP Server

What it does: Gives your AI assistant secure, controlled read and write access to folders on your local machine or a shared network drive — no custom integration required. You define exactly which directories are accessible, and the server enforces those boundaries so the model cannot touch anything outside the permitted paths.

Best for: Operations managers, HR business partners, and finance analysts who regularly work with large volumes of documents (policy files, reporting packages, employee records drafts) and want to ask questions, extract data, or generate summaries without copying and pasting content into a chat window one file at a time.

How to deploy:

  1. Install the Filesystem MCP server via the MCP Registry (registry.modelcontextprotocol.io) and point it at one or more approved local or network directories — for example, your quarterly reports folder or your HR policy library.
  2. Connect the server to your MCP-compatible AI client (Claude Desktop is the most common starting point) by adding the server entry to your client’s MCP configuration file, specifying only the folder paths you want exposed.
  3. Restart your AI client and verify access by asking it to list the files in the directory — then start issuing natural-language queries like “Summarize all files modified this week” or “Find every document that mentions the word indemnification.”

Why it matters: The single biggest bottleneck in knowledge-worker AI adoption is the copy-paste tax — the manual effort of moving content from a file system into a chat interface. The Filesystem MCP eliminates that entirely for document-heavy workflows. An HR team can ask their AI to cross-reference a new policy draft against every existing policy in the library before it goes to legal review. A finance team can ask for variances across twelve monthly reports without opening a single spreadsheet. Because access controls are set at deployment time and respected at runtime, this also satisfies the most common IT objection: that the model will “see everything.” It only sees what you tell it to see.

Also worth trying

The Meeting-to-Action Converter (Prompt). Paste a raw meeting transcript or even rough personal notes and ask: “You are a chief of staff. Extract: (1) all decisions made, (2) all action items with owner and deadline as stated or implied, (3) any open questions requiring follow-up, and (4) a three-sentence summary for stakeholders who were not present. Format as a structured memo. Transcript: [PASTE HERE].” Operations and project management teams report this alone saves 20–30 minutes per meeting. Works best when you paste a verbatim transcript from a tool like Otter.ai or Zoom rather than edited notes.

Memory MCP Server. This reference server maintains a persistent knowledge graph that survives across separate conversations. For HR and operations leaders who use an AI assistant daily, it means you can tell the model once that your org uses a specific performance review framework, your fiscal year ends in March, or your preferred vendor for a category is already contracted — and it will remember that context in every future session. Deploy it from the MCP Registry, connect it to Claude Desktop, and start by having the AI interview you for five minutes about your role, your team structure, and your recurring workflows. The payoff compounds over weeks.

The Stakeholder Communication Calibrator (Prompt). Built for marketing and internal comms teams, this prompt takes a technical or dense source document and rewrites it for a specified audience at a specified stakes level. Use it as: “Rewrite the following [DOCUMENT TYPE] for an audience of [AUDIENCE, e.g., ‘non-technical board members’] who will use this to make a [DECISION TYPE] decision. Prioritize clarity over completeness. Maximum length: [WORD COUNT]. Flag any nuance you had to omit. Source: [PASTE HERE].” Particularly effective for translating IT security incident reports into executive briefings or converting actuarial summaries into client-facing language.

← All prompts