A Claude Skill for Consulting-Grade Problem Solving

Quick answer: which route fits your question?

There is no universal right answer, only the right one for your stakes and your deadline. Match the route below to how much judgment and time you actually have, not to whichever tool happens to be open.

For most professionals who already use Claude, the free 8-stage Claude skill for problem solving on this page is the strongest general-purpose route: it forces a decision question, enforces MECE logic through a dedicated stage, and hands off a challenge-tested recommendation rather than raw prose. If you only need the issue tree itself, the narrower MECE Issue Tree skill for Claude is faster. If the stakes justify a staffed team and a multi-week timeline, a top-tier strategy engagement still produces the most complete deliverable. Plain prompting without any skill works for a first pass, but expect more rework later.

How we chose what belongs in this method

This method earned its stages the same way a consulting deliverable earns a slide: it had to survive being questioned. The alternative is real and it is noisy: as of June 2026, roughly 40 skills-compatible products list on Anthropic's open Agent Skills standard, and community catalogs such as SkillsMP index close to 1.9 million public skills, scraped from GitHub and of wildly uneven quality. Four rules decided what stayed in this one, and a fifth decided what to leave out.

• Has to work from an undefined question. Every stage assumes you have not yet decided what the real question is; techniques that assume a pre-scoped brief did not make the cut.
• Has to force MECE, not just describe it. Stage 2 tests every branch pair for overlap and gaps before analysis proceeds, instead of leaving MECE as an instruction Claude might or might not follow.
• Has to separate thinking from rendering. The skill stops at approved content. Slide design, chart building, and file rendering are explicitly out of scope and handed to other tools.
• Has to survive a hostile question. Every recommendation runs through an independent challenge stage before packaging, the same discipline a partner review is meant to provide.
• What we left out on purpose. Data gathering, financial modelling, and market sizing are not automated. The method assumes you already have, or can get, the underlying facts.

Key takeaways

• A route is only as good as the question you hand it. If you cannot yet state the decision in one sentence, no skill, prompt, or engagement will produce a defensible answer; write the question first and choose the route second.
• MECE has to be enforced by something, not just invoked. Plain prompting and a hand-drawn tree both rely on the person doing them remembering to check for gaps and overlaps; a skill that tests every branch removes that failure point.
• Know whether you are buying a specification or a deliverable. A skill can hand you approved content ready for a builder to render; only a full engagement hands you the finished deck or memo itself.
• Speed and completeness trade against each other on every route here. The fastest option needs the most rework afterward; the most complete option needs the most lead time before it starts.
• A recommendation that has never been challenged is a draft, not an answer. Routes without a built-in review stage need you to add one yourself before it goes in front of anyone senior.

Five ways to get to a recommendation, compared

Read the table against the five rules above, not against whichever option you already have open. The route that wins each column is not always this skill, and the category leader for judgment and completeness is not an AI tool at all.

All of this runs on Anthropic's open Agent Skills standard, including the narrower MECE Issue Tree skill for Claude. For the primary reference on how any Claude skill actually loads, see Anthropic's own engineering guide to Agent Skills.

This Skill: Who Is It Built For?

This is for a professional who already uses Claude, has a real ambiguous business question, and wants a repeatable method rather than a one-off clever prompt. It behaves like a sharp engagement associate: framing the question, building the logic, drafting the analyses, writing the recommendation, while you stay the partner setting direction, adding context, and approving or revising each stage's output.

Two gates run at every stage: an input gate where you supply context, constraints, and preferences, and a review gate where you approve the deliverable or send it back before the workflow continues. That structure matters more than it sounds. SkillsBench, an independent benchmark that evaluated 47,150 publicly available Claude skills, found an average quality score of just 6.2 out of 12, while curated, high-quality skills lifted agent task pass rates by 16.2 percentage points on average. Downloading a random public skill for a task like this one is a real gamble; this one is built to the stricter bar described in How We Chose above, and it is free to inspect before you install it.

The middle of the method is where most of the judgment lives. Stage 5 does not just run analyses; it labels each finding with a confidence level, so a hunch and a tested fact never look the same on the page by the time Stage 6 pulls them into one point of view. That labelling is what a reviewer, or a partner, actually checks first.

The trade-off: it only produces approved written content: a recommendation, a storyline, supporting analysis. It never produces a rendered slide, chart, or file. Pair it with a downstream builder skill, or a tool like Oria, once Stage 8 is approved.

The 8 Stages You Get

Each stage runs through the two gates described above: an input gate where you provide context, judgment, constraints, and preferences, and a review gate where you approve or revise the deliverable before the workflow continues. The result is a clean trail from the original question to the final recommendation.

Stages 1 to 4 sharpen the question and the plan before any analysis runs. Stages 5 to 8 carry the answer through to an approved recommendation and packaged content.

Plain Claude or ChatGPT Prompting: Who Is It Built For?

This is for a genuinely low-stakes question, a first pass before a bigger process, or someone testing whether AI-assisted structuring is worth adopting at all. Open a chat, describe the problem, and ask Claude or ChatGPT to think it through. No download, no setup, and it works today.

The cost shows up later. Nothing forces a decision question, nothing checks that your branches are mutually exclusive and collectively exhaustive, and nothing challenges the recommendation before you act on it. All of that depends entirely on the discipline of whoever is typing. That gap is exactly what structured methods exist to close. Paul Nutt, professor of management sciences at Ohio State University's Fisher College of Business, studied more than 400 real decisions made by managers across the United States, Canada, and Europe over more than 20 years and found that about half of business decisions end in failure. As strategy writer Roger Martin puts it, "Strategy is the act of making a new set of choices designed to solve a problem." An unframed problem produces choices that solve the wrong thing.

The trade-off: the lowest possible barrier to starting, and the highest amount of rework if the question turns out to matter. Fine for a first pass; risky as the sole basis for something that has to survive a room of senior stakeholders.

The MECE Issue Tree Skill for Claude: Who Is It Built For?

This is for someone who has already framed the decision question and needs one thing done well: a tested, MECE issue tree for that specific question, without the seven stages around it. The MECE Issue Tree skill for Claude, also free on this site, is narrower by design: a two-level decomposition method with branch tests, diagnostic questions, and assumption checks, built for exactly that one job.

That narrowness is the point, not a limitation. If your team already has a hypothesis process, an analysis plan, and a house deck format, and the only gap is that trees keep coming back with overlapping branches, installing an eight-stage workflow to fix a one-stage problem is overkill. For the underlying MECE test itself, independent of which skill or format applies it, see the MECE examples and meaning guide, and for ways to actually draw and present the resulting tree, see the issue tree guide on this site.

The trade-off: it decomposes one question fast, but stops there: no hypotheses, no analysis plan, no recommendation. Everything after the tree is still on you.

A Top-Tier Strategy Engagement: Who Is It Built For?

This is for a decision where the cost of being wrong dwarfs the cost of the engagement itself: a major market entry, a turnaround, a board-level restructuring. A top-tier strategy firm brings senior judgment, a staffed team, and the same MECE and hypothesis-led practice this entire method is modelled on, refined over decades of exactly this kind of question.

It also brings a proposal cycle, weeks of lead time before the first workstream starts, and staffing costs measured in consultant-weeks rather than a single evening prompt. None of that is a flaw; it is what the format is actually for. The trade normally makes sense only when the decision is genuinely board-level, the underlying facts are not yet gathered, or the recommendation needs a name and a track record behind it that a downloaded skill cannot supply.

The trade-off: the most complete and most defensible route on this page, and by a wide margin the slowest and most expensive. Wrong choice for a decision that has to ship this week; arguably the only right choice for one that has to survive a board.

If you need something else, go here

Many readers land on a query like this one needing something narrower, or something bigger, than the full eight-stage method. Find your row and go straight there.

Setup Guide

1. Step 1

   Download the skill and open Settings

   Download the skill (.zip)

   Download and unzip the single skill bundle. It contains one top-level SKILL.md and eight supporting files in references/. In Claude, open your profile menu and select Settings.

   

2. Step 2

   Open Skills and start an upload

   Skills are available on Claude's Free, Pro, Max, Team, and Enterprise plans, but they need Code execution turned on first, under Settings, then Capabilities. With that on, select Skills under Customize. Open Add and choose Upload a skill.

   

3. Step 3

   Upload the skill file

   Drag the extracted top-level SKILL.md into the upload window, or use the file picker. Keep the downloaded references/ folder because it contains the full ordered method for Claude Code.

   

4. Step 4

   Confirm the skill was added

   Return to the Skills list and check that the problem-solving skill appears. Select its name to open it.

   

5. Step 5

   Review and turn on the skill

   Review the skill name, description, and instructions, then make sure its switch is on. A blue switch means the skill is enabled; sharing it is optional.

   

Why it works

The principles under the hood

Every stage is grounded in a specific working principle. No generic "think strategically" prompts. Each principle is named, applied at the right gate, and traceable from the recommendation back to the original question.

Frequently asked questions