Guide · Agent skill
How to use the Deal Desk Review agent skill (deal-desk)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The deal-desk skill is a set of instructions and scripts that lets an AI agent review one inbound deal before it closes: it scores margin and risk, routes discount approval to a named human, and flags risky contract terms in a redlined MSA. It is aimed at revenue operations and deal desk staff, sales leaders who want a consistent approval rule, and founders triaging a customer's redlines. After this guide you will know what the skill does, how to install it, how to phrase requests so it triggers, and what to check before relying on its output.
Deal Desk Reviewalirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the deal-desk skill ZIP from this page.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. If you use the Claude apps, upload the skill under Settings → Capabilities → Skills.
- 4Prepare the deal JSON. Fill in assets/deal_intake_template.md with ARR, discount, payment terms and term details, and save the JSON inputs.
- 5Ask for a review in plain language. Describe the deal, the discount or the redlined MSA, and ask who must approve.
- 6Read the packet. Check the score, verdict, approver chain and redline findings, then send the recommendation to the named human approvers.
What agent skills are and what deal-desk makes the agent do
An agent skill is a folder with a SKILL.md file, reference documents and optional scripts that an AI agent such as Claude Code loads when a request matches the skill's description. The deal-desk skill is designed for the moment between a sales team asking for a discount and a CFO, CRO or General Counsel signing off.
The skill never approves anything. Every output is a numeric scorecard plus a routing recommendation to a named human approver, such as a Sales Director, VP Sales, CFO, CRO or General Counsel.
- Scores a deal from 0 to 100 across five dimensions (margin, risk, strategic value, commercial fit, term shape) and gives one verdict: APPROVE, REVIEW, ESCALATE or DECLINE.
- Routes a discount request to a named approver chain with estimated cycle days, using deal_scorer.py and discount_approval_router.py.
- Scans structured deal terms for 10 contract landmines such as uncapped indemnity, MFN, perpetual license-back and missing DPA, with suggested counter-language, using terms_redliner.py.
- Assembles the three outputs into a deal-desk review packet that is a recommendation, not an approval.
Install the deal-desk skill from the ZIP
Download the ZIP for the skill from this page, then unzip it into one of two locations. Use ~/.claude/skills/ to make the skill available across your projects, or use a project's .claude/skills/ folder to limit it to one project.
If you use the Claude apps instead of Claude Code, you can upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/ so the deal-desk folder sits there.
- Project install: unzip into the .claude/skills/ folder inside your project.
- Claude apps: upload the skill under Settings → Capabilities → Skills.
Check the requirements before you start
The scripts use only the Python 3 standard library, so there are no packages to install beyond Python 3 itself. You also need a deal JSON file containing ARR, discount, payment terms and term details. The skill's intake template, assets/deal_intake_template.md, is the source for that file and takes about 20 minutes to fill out.
You can optionally pass your own discount policy thresholds in the input JSON. Without them, the built-in bands are generic industry defaults and routing may not match your business.
- Python 3 (scripts use only the standard library).
- A deal JSON file with ARR, discount, payment terms and term details, filled in from assets/deal_intake_template.md.
- Optional: your own discount policy thresholds in the input JSON.
- The skill runs scripts but does not use the network.
Trigger the skill with natural requests
The skill loads when your request matches its description. Describe the situation in plain terms: a discount above the rep's authority, a customer's redlined MSA, a need to quantify margin or payment shape, or a need to know who must approve.
Name the deal and attach or paste the deal details so the agent can locate the intake data. The skill is for one specific deal at a time.
- "Review deal ACME-2026-Q2-117. The AE wants 30% off list. Score it and tell me who has to approve."
- "The customer redlined our MSA. Triage the terms and list what to push back on before it goes to counsel."
- "Quantify the post-discount margin and the multi-year payment shape for this renewal expansion."
- "Assemble the deal-desk review packet for this inbound deal with the named approver chain."
Read and use the output
Read the composite score, the per-dimension breakdown and the verdict together. The skill's guidance is that critical signals override the composite: a high score with an uncapped indemnity finding is still a DECLINE. The approver chain should be shown in full, and modifiers such as the enterprise floor or SMB fast-lane are surfaced explicitly rather than hidden.
Treat every verdict, including APPROVE, as a recommendation that names the people who must sign.
- Check the composite score and the five dimension scores.
- Check the verdict and the named approver chain.
- Review each redline finding with its severity (CRITICAL, HIGH, MEDIUM, LOW) and counter-language before sending anything to counsel.
Safety notes: scripts and what the skill does not do
The skill runs Python scripts locally and, according to its metadata, does not use the network. Review the scripts in the scripts/ folder before running them on confidential deal data, as you would with any skill that executes code.
The terms scanner works on a structured JSON summary, not on full contract prose. It is a triage tool and is not a substitute for General Counsel review. The skill also does not draft proposals or contracts, and it does not set your discount policy.
- Scripts run locally; the skill's metadata lists no network use.
- Terms scanner input is a structured terms JSON, not contract text.
- Every HIGH or CRITICAL finding should be reviewed by named counsel before signing.
- Override the default discount bands with your own matrix, or routing may be wrong for your business.
Tips for better results
Fill out a new intake for each re-pricing or term change rather than editing the old one, since the template is the single source of truth for the deal. Keep your discount policy in the input JSON so routing follows your rules rather than generic defaults.
Use the skill for routing, not for choosing prices. It applies a discount the AE or customer has already proposed; pricing strategy belongs to a separate pricing skill.
- Create a new intake for every re-pricing or term change.
- Pass your own policy_thresholds in the input JSON.
- Use the scripts' --help and --sample options to see expected input shapes.
- Use the scripts' --output option to choose human-readable or JSON output.
What people use it for
- Scoring an inbound discount request above AE authority before it goes to a Sales Director or VP Sales.
- Triaging a customer's redlined MSA to flag uncapped indemnity, MFN or missing DPA before sending it to counsel.
- Quantifying post-discount margin and multi-year payment shape for a renewal expansion.
- Building a consistent, documented approval routing rule for a deal desk queue.
Frequently asked questions
How do I install the deal-desk agent skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
What are the requirements for deal-desk?
You need Python 3, since the scripts use only the standard library. You also need a deal JSON file with ARR, discount, payment terms and term details, filled in from the intake template. Your own discount policy thresholds are optional.
Is deal-desk free to use and what is its licence?
The skill is listed under the MIT licence. The source does not state any price, so check the directory page for current terms.
Does deal-desk approve discounts automatically?
No. The skill never auto-approves. Every output is a numeric scorecard and a routing recommendation to a named human approver.
Can deal-desk review a full contract or draft a proposal?
No. The terms scanner works on a structured terms JSON, not full contract prose, and the skill does not author proposals or contracts. For drafting, use a contract-and-proposal writing skill; for full legal review, use General Counsel review.
What should I do if the default discount bands do not match my company?
Pass your own discount policy thresholds in the input JSON. The built-in bands are generic industry defaults, so without your own matrix routing may be wrong for your business.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Deal-desk suits revenue operations staff, sales leaders wanting a documented approval rule, and founders triaging a redlined MSA before counsel. It is a poor fit if you need to draft contracts from scratch or get a full legal review of contract prose, and you should use your own discount matrix rather than the generic defaults.
Download the skill