Every Claude Skill for Procurement Teams (A Live Inventory)

We spent the last quarter cataloguing every Claude skill a procurement team can actually put to work. Not the demo-day list. The ones we run against real RFPs, real supplier responses, and real category reviews, week after week. The list is shorter than the marketing suggests and more useful than most procurement leaders expect.

This is that inventory. We have grouped the skills by the procurement workflow they serve, added an honest note on what each one does well and where it breaks, and flagged which ones you can pull off a shelf versus which ones you have to build yourself. We will keep this page current as Anthropic ships new skills and as we build more of our own.

If you are still deciding whether Claude belongs in your procurement stack at all, start with our Claude Cowork playbook for procurement teams. This piece assumes you are past that question and want to know exactly what to load.

What a Claude Skill Actually Is (And Why Procurement Should Care)

A Claude skill is a small, reusable package of instructions that Claude loads only when it is relevant. Under the hood it is a folder with a short instruction file (the skill tells Claude its own name, when to use it, and how), plus any templates or scripts the task needs. Claude reads the one-line description of every available skill, and when your request matches, it pulls in the full instructions on demand. Anthropic calls this progressive disclosure. In plain terms, the model stays lightweight until a skill is needed, then it behaves like it has read your playbook.

That last part is why procurement should care. A prompt is a one-time instruction. A skill is a standing capability. The difference shows up the third time you run a task. With prompting, someone has to remember the good prompt and paste it correctly. With a skill, the RFP scoring rubric, the redline standards, the category framework you use are already baked in. The junior analyst gets the same output structure as the category lead.

Skills run wherever Claude runs: the Claude apps, Claude Code, and through the API for teams that want them inside their own tools. Anthropic's own overview of Agent Skills covers the mechanics. For a procurement team, the practical implication is simple. Instead of teaching every person to prompt well, you teach Claude your workflow once and everyone inherits it.

Skills for Sourcing and RFx

This is where procurement teams get the fastest payback, because the work is document-heavy and rule-bound, which is exactly what skills are good at. RFx is the umbrella term for the request documents you send suppliers: an RFI to gather information, an RFP to solicit proposals, an RFQ to get priced quotes.

The RFP generator skill. Point it at a scope of work or a rough brief and it drafts a structured RFP, RFI, or RFQ: sections, requirement tables, evaluation criteria, and a scoring framework. The version we run turns a two-page brief into a first-draft tender in minutes rather than the day or two it takes to assemble one from an old template. It does not replace category judgment. It removes the blank-page tax. We wrote up a live example of this in how we use Claude for RFP work.

The RFP response evaluator skill. The mirror image. Drop a folder of supplier responses and it scores each one section by section against the criteria you set, then produces a side-by-side comparison and a ranked shortlist with the reasoning shown. The honest limitation: it is only as good as the rubric you give it. Hand it vague criteria and you get confident, vague scores. Hand it a weighted rubric with clear thresholds and it becomes the most patient evaluator on your team.

The implication for sourcing leads is that the two skills chain. Generate the RFP with the criteria embedded, then evaluate responses against those same criteria. The rubric travels from the send side to the score side, which removes the classic problem of scoring suppliers against standards you never actually published.

Skills for Category Strategy and Negotiation

These skills carry more strategic weight, and they are the ones we are most careful about, because the output feeds decisions rather than documents.

The category strategy builder skill. It structures a category plan using established frameworks, including Kraljic segmentation (the two-by-two that sorts spend by supply risk and profit impact, so you know which categories to partner on and which to simply source competitively). Feed it your spend data and market context and it produces a first-pass segmentation, a supply-market read, and a savings-opportunity view. We treat its output as a strong draft that a category manager sharpens, never as the final strategy. The framework discipline is the value. It stops the analysis from being an unstructured opinion dressed up in a slide.

The negotiation playbook generator skill. It builds a structured negotiation plan: BATNA analysis (your best alternative to a negotiated agreement, the walk-away that sets your real leverage), the sequence of levers to pull, and counter-proposal language. For a renewal or a competitive re-source, it gets a category lead to a defensible plan far faster than a blank document does. The limit is judgment about the relationship. The skill does not know that this supplier saved your line during a shortage two years ago and has earned a softer stance. You do.

The pattern across both skills is the same. They enforce a framework and generate the first ninety percent. The last ten percent, the part that reflects context only your team holds, stays human. That is the correct division of labour, and any skill that claims to remove the last ten percent is overselling.

The Document Skills That Do the Unglamorous Work

The least discussed skills are the ones procurement teams end up using most, because so much procurement output is a document someone else has to open in Microsoft Office or a PDF reader.

• The .docx skill turns a drafted contract, policy, or memo into a properly formatted Word document, with headings, tables, and tracked changes intact. This matters when legal expects a redline, not a chat transcript.
• The .xlsx skill builds and reads spreadsheets. Supplier scorecards, savings trackers, and should-cost models come out as real Excel files your finance partner can pivot, not as a table pasted into an email.
• The .pptx skill assembles slide decks. The quarterly business review or the board-approval deck goes from analysis to formatted slides without a manual rebuild.
• The .pdf skill reads, splits, merges, and fills PDFs. Useful for pulling terms out of a signed contract or filling a supplier onboarding form at volume.

None of this is glamorous. All of it is the reason the AI output actually gets used. A brilliant analysis that lives in a chat window and has to be rebuilt by hand before anyone can present it is not a time saving. It is a second draft. The document skills close that gap, and they are the first ones we install for any team that runs Claude in a real approval chain.

The Skills No Marketplace Will Give You (Build These Yourself)

The shelf skills get you a long way. The differentiated value comes from the skills you build for your own organisation, because they encode things no vendor could know.

Building one is more approachable than most procurement leaders assume. A skill is, at its simplest, a written description of when to use it plus the instructions to follow. You do not need to code to write down how your team scores a supplier or structures a savings memo. The skill file is closer to a good standard operating procedure than to software. Anthropic's Agent Skills documentation lays out the format, and the honest truth is that the hard part is not the syntax. It is deciding what your standard actually is.

The custom skills worth building first are the ones that turn tribal knowledge into a repeatable capability:

• Your supplier risk scoring model, so every risk review weighs the same factors the same way.
• Your redline standard, so the clauses your legal team always fights over get flagged before the contract reaches them.
• Your intake triage, so incoming requests get routed and pre-scoped against your own thresholds.
• Your reporting format, so the monthly procurement update comes out in the structure your CFO already expects.

We build these with clients constantly, and we have built them for our own operation. One example is a skill that audits and rewrites cold outreach against a framework we trust, so the output is on-standard without a human editor in the loop every time. The point is not that skill in particular. The point is that the skills carrying the most weight are the ones only your team could have written.

The Honest Limitations of Skills

Skills are the most useful Claude feature we have deployed in procurement. They are also routinely oversold, so here is the balancing view.

A skill does not fix bad data. If your spend is uncategorised and your supplier master is a mess, a category strategy skill will produce a well-structured plan built on unreliable inputs. The structure hides the rot. We have watched teams mistake a tidy output for a trustworthy one. Fixing the data is still the first project, as we argue in our guide to implementing AI in procurement.

A skill also does not manage adoption. Installing the RFP evaluator does not mean your team will use it, especially if the incumbent process is comfortable and the new one requires trusting a machine's shortlist. The skill is the easy part. Getting people to change how they work is the hard part, and it is where most procurement AI effort actually gets stuck.

And a skill inherits your judgment, including the flawed parts. If your scoring rubric overweights price and underweights delivery reliability, the skill will do that faithfully at scale. The consistency that makes skills valuable also means a bad standard gets applied consistently. Build the standard carefully, because the skill will not second-guess it.

Where to Start

If you want Claude skills working in your procurement function this quarter, do not start by installing everything. Start with one workflow that is document-heavy, rule-bound, and run often. RFP evaluation is usually the best first candidate, because the rules are explicit and the payback is visible within a single sourcing cycle.

Install one shelf skill, run it against a live RFx, and compare its output to what your team produced by hand. Then write down the one thing it got wrong about your standards, and turn that correction into your first custom skill. That loop, install then correct then encode, is how a procurement team goes from prompting Claude to running it. Our procurement AI prompt library is a good source of raw material to convert into skills once you know which workflows to target.

We will keep adding to this inventory as the skill library grows and as we build more with clients. The specific skills will change. The principle will not: the shelf skills get you started, and the skills you build yourself are the ones that become a real advantage.