Guide · Agent skill
How to use the Commercial Discount Policy Designer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The commercial-policy skill is an agent skill: a folder of instructions, reference material and Python scripts that an AI coding or chat agent can load when a task matches its description. This one helps a Head of Commercial, Head of Deal Desk, VP Sales or RevOps lead write the company's rules for how far list price may be discounted, who approves each level, and how unusual exceptions are handled. After this guide you will know how to install it, how to ask for it in plain language, and what its scripts produce.
Commercial Discount Policy Designeralirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the commercial-policy ZIP from this page.
- 2Install the skill. Unzip into ~/.claude/skills/ or a project's .claude/skills/, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Prepare deal data. Export about four quarters of closed-won and closed-lost deals as JSON with the required fields.
- 4Describe the policy problem. Ask the agent in plain words to design or revise the discount policy, naming the policy moment rather than one deal.
- 5Build the matrix. Run discount_matrix_builder.py with your intake file and a profile, and review THIN cells.
- 6Design exceptions. Run exception_router.py --sample and codify the severity bands and compensating commitments in your policy document.
- 7Lint and publish. Run policy_linter.py on the matrix, resolve all BLOCKER findings, then publish a versioned matrix.
What the commercial-policy skill makes the agent do
The skill is for designing policy, not approving individual deals. Its stated job is to produce the written rules that Deal Desk and account executives work under: a discount matrix, an exception flow and a lint report on that matrix.
The matrix is built across four dimensions: ARR band, term length, payment terms and strategic value tier. Each cell carries an approved discount band, an approver tier (AE, Manager, Director, VP or CFO), a margin floor and observed win-rate and NRR figures from past deals.
- discount_matrix_builder.py builds the 4-dimensional matrix from your deal data and a chosen industry profile (saas, enterprise-software, api, marketplace or services).
- exception_router.py routes requests that fall outside the matrix through the named approver chain and attaches required compensating commitments.
- policy_linter.py checks the matrix for governance defects such as contradictions, coverage gaps and cliff edges, and ranks findings as BLOCKER, MAJOR or MINOR.
Install the skill from the ZIP download
Download the ZIP for commercial-policy from this page. Then install it using one of two routes, depending on where you work.
For Claude Code, unzip the file into your personal skills folder ~/.claude/skills/ or into a project's .claude/skills/ folder. For the Claude apps, upload the skill under Settings → Capabilities → Skills.
- Personal use across projects: unzip into ~/.claude/skills/.
- Single project, shared with a team: unzip into the project's .claude/skills/ folder.
- Claude apps: upload through Settings → Capabilities → Skills.
Check what you need before you start
The scripts are stdlib-only Python 3, so there are no packages to install beyond Python 3 itself. The builder needs exported CRM deal data in JSON. The skill's own intake template lists the fields it expects per deal: arr, discount_pct, term_months, payment_terms_days, strategic_value, win_lost and nrr_12mo, covering roughly the last four quarters of closed-won and closed-lost deals.
The skill runs scripts, and it does not use the network.
- Python 3 installed locally.
- A JSON export of closed-won and closed-lost deals with the fields listed above.
- Named owners for CFO sign-off and CRO or VP Sales sign-off, as captured in the intake template.
Trigger the skill with natural requests
You do not need special syntax. Describe the policy problem in plain words and mention the policy moment rather than a single deal. The skill is meant for requests like designing a first discount policy, rebuilding a drifted matrix, or setting approval thresholds.
Requests that belong elsewhere should be sent to other skills. The source says per-deal approval belongs to deal-desk, and list price or pricing model choice belongs to pricing-strategist.
- "Help me design our first formal discount policy for SaaS, with approver tiers by discount level."
- "Our discount matrix is 14 months old and reps keep citing last quarter's exceptions. Help me rebuild it from our closed deals."
- "Lint this discount matrix for contradictions, cliff edges and gaps before we publish it to AEs."
- "Design the exception flow for asks 5 to 10 points over the band, with compensating commitments."
Run the scripts for matrix, exceptions and lint
Each script supports --help and --sample, and accepts --input and --output with markdown or json. Start with --sample to see the structure before using real data.
The builder takes your intake JSON and a profile. The linter takes the matrix output. The skill's workflow asks you to resolve every BLOCKER before publishing the matrix to AEs.
- Cells with fewer than 5 observed deals are flagged THIN.
- Exception severity bands are 0-5, 5-10, 10-20 and 20+ points over the band.
- The router flags precedent risk when 3 or more similar exceptions land in the trailing quarter.
python scripts/discount_matrix_builder.py --input policy_intake.json --profile saas
python scripts/exception_router.py --sample
python scripts/policy_linter.py --input matrix.jsonReview and republish each quarter
Publish the matrix as a versioned artifact, and version the intake the same way. The intake template says re-pricings or major revisions should create a new intake rather than edit the old one.
Re-run the builder and linter each quarter against the rolling four-quarter deal corpus. Cells where observed NRR is below target_nrr are flagged for review.
- Treat a cell with observed NRR below target as a signal to review the band, not to discount deeper.
- Keep the constraints separate: the CFO owns the margin floor, and the CRO or Head of Deal Desk owns the band cap.
What people use it for
- A Head of Commercial writing a company's first formal discount policy with approver tiers.
- A RevOps team rebuilding a discount matrix that reps have overtaken with precedent, such as citing past approved discounts.
- A CFO or exec team asking why discounts are at their current levels and needing a data-backed rationale.
- Checking a draft matrix for approver inversions, band inversions, coverage gaps and cliff edges before it reaches AEs.
- Defining the exception flow, including which approver is needed and which compensating commitments apply at each severity.
Frequently asked questions
How do I install the commercial-policy 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 for a single project. In the Claude apps, upload it under Settings → Capabilities → Skills.
What do I need to run the commercial-policy skill?
You need Python 3, since the scripts are stdlib-only, and exported CRM deal data in JSON for the matrix builder. The deal data should include ARR, discount, term, payment terms, strategic value, win or loss and NRR. The skill does not use the network.
Is the commercial-policy skill free to use?
The skill is published under the MIT licence, as stated on this page. The source does not mention pricing, so check the licence text in the download for the terms.
Does the skill approve individual deals?
No. The source says the skill designs policy and never says approve. Per-deal approval belongs to a separate deal-desk skill, and pricing model or list price selection belongs to pricing-strategist.
What if I don't have closed-won and closed-lost deal data?
The skill is not suited to that situation. The matrix bands and win-rate figures would be unsupported without real closed-won and closed-lost deals, so build the data first.
Do the scripts send my data anywhere?
The source states the skill uses no network access, and the scripts are stdlib-only Python. They run locally on the files you give them, and the script runs are the part to review for safety before use.
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
Use commercial-policy if you are setting or rebuilding a company's discount rules and have real closed deal data to back them, especially if you are writing a first formal policy or fighting precedent-driven discounting. Skip it if you need to approve a single deal, choose a pricing model, or lack historical deal records, since the source points those tasks to other skills or assumes data you do not have.
Download the skill