OPERATOR SKILL
Find out whether your plan is actually a strategy
An AI skill that reads a plan, deck, or brief, extracts its diagnosis, guiding policy, and actions, names the ones that are missing, and writes the replacement sentences. It applies Richard Rumelt's Good Strategy/Bad Strategy framework in the AI tool you already use.
- Instant access
- Lifetime updates
- Secure purchase
THE OUTPUT
One audit, with your own words as the evidence
Point it at a plan and you get one written audit. Every section has a fixed job:
- Verdict: one of four defined ratings, so two audits can be compared
- Kernel scorecard: diagnosis, guiding policy, and coherent actions, each rated Present, Weak, or Missing, each with a quote and where it came from
- Four-hallmark screen: fluff, failure to face the challenge, goals mistaken for strategy, bad objectives, with one line of evidence each
- What is working, limited to specific moves it can quote
- Findings ranked by consequence, not by order of appearance
- How to make it good strategy: the repairs, upstream first
- What the plan is choosing not to do, including who will object
- First test: the cheapest observation that would show the diagnosis is wrong
- Open questions, and the limits of the audit itself
A full audit of a ten to twenty page plan runs about 900 to 1,600 words. Ask for a quick screen instead and you get the scorecard, the verdict, and the top three fixes in under 400.
THE PROBLEM
Everyone agreed with the plan, and nobody can say what it overcomes
Most documents called strategy are not strategies. They have a vision, a situation section, a target table, and a budget, and they read perfectly well.
What they do not have is a sentence explaining why the thing is hard. Without it, every objective is a wish about the result, and the plan cannot be argued with, improved, or rejected.
That failure is an omission rather than an error, which is why it survives review. The document is detailed, internally consistent, and unanimously approved. Then the targets are missed, and execution gets the blame for failing to deliver something that never had a lever.
HOW IT WORKS
Extract the kernel, screen for the four hallmarks, then repair
The audit runs the same phases every time:
- Read all of it first, including appendices and slide notes, because the diagnosis is often buried in exactly those places
- Extract the kernel: quote the diagnosis, the guiding policy, and the coherent actions, each with its location, and rate each one Present, Weak, or Missing
- Run the four-hallmark screen in a fixed order, with one line of quoted evidence for each
- Apply targeted lenses, chosen from fifteen by the symptoms your document actually shows
- Issue the verdict and rank the findings by consequence
- Prescribe the repairs, diagnosis first, then policy, then objectives and actions
- Self-check against eight fairness rules before it delivers
If an element cannot be quoted, it is rated Missing. Saying so plainly is the finding.
What's included
What you get
The Strategy Audit skill
The seven-phase workflow, three audit modes, the four verdict definitions, eight fairness rules the audit checks itself against, and the symptom-to-lens routing table.
Kernel and hallmarks reference
How to extract a diagnosis, a guiding policy, and a coherent action set, the failure signature of each, and the four hallmarks of bad strategy with their quick tests.
Diagnostic lenses reference
Fifteen targeted lenses, each with the question to put to the document, the red flag, the good sign, and the direction of the fix.
Repair playbook
Six repair rules, a nine-step repair sequence, before and after rewrites for each failure type, templates for hypotheses and policies, and a question bank for the plan's owner.
Evidence bank
Compact cases grouped by the finding they support, each with an honest caution about what it does and does not show, plus the limits that apply to the framework as a whole.
Audit report template
The ten-section report structure with guidance for each section and the formatting rules, including the shortened version used for a quick screen.
Worked example
An invented plan and a complete audit of it, showing the expected depth, the evidence style, and how specific a repair is meant to be.
THE REPAIR
It writes the replacement sentences
An audit that only criticizes is half a deliverable. Every finding comes back paired with something the owner can act on this week:
- Candidate diagnoses as hypotheses, each with the facts that would make it true, the one observation that would refute it, and what it rules in and out
- Two candidate guiding policies, each a single sentence, each with five specific actions it excludes
- One proximate objective, dated and measurable, plus the ambiguity it resolves for the team
- Three to six coherent actions, each with a first step and an owner, coordinated around one named challenge
- A list of what to stop, including who will be unhappy about it
It will not invent your facts. Where a reading depends on something your document does not contain, it names the missing fact and a cheap way to get it, such as a call log review or a cohort cut, instead of filling the gap with a plausible guess.
Use cases
Where it fits
Pressure-testing a plan before you fund it
Run it on the plan that is about to set next year's budget. You get the kernel scorecard, the findings ranked by consequence, and the cheapest test that would show the thinking is wrong.
Reviewing a plan from a team or an agency
Judge a document in a domain you do not know well. The test is structural, so the findings rest on quoted evidence instead of on your read of their industry.
Cutting an objective list down to what matters
Turn a long unprioritized list into one pivotal objective, the items it enables, the ones that act as constraints, and the ones being dropped, each with the reason.
Rebuilding a plan that keeps missing its targets
When execution gets blamed every quarter, rebuild mode produces a diagnosis, a guiding policy, a proximate objective, and a coordinated action set with first steps and owners.
Capabilities
What it does for you
Evidence, not opinion
Every finding quotes your document and says where the quote came from. You argue with the evidence rather than with someone's taste.
A verdict you can compare
Four ratings with written definitions: good strategy, promising but incomplete, bad strategy, or not a strategy. Last quarter's audit and this one mean the same thing.
Fifteen lenses, chosen by symptom
Lenses are selected by what your document actually shows, using a routing table. Running all fifteen on one plan is treated as noise, not thoroughness.
The repair is half the deliverable
Every finding comes back with the replacement sentence written out, two candidate policies with the actions each one excludes, and a named list of what to stop doing.
It never invents your facts
Candidate diagnoses are labeled hypotheses with the facts that would make them true and the observation that would refute them. Missing facts are named, not guessed.
It says when the framework does not apply
A plan facing a hard obstacle honestly is rated promising, not bad. When the situation is routine and nothing is broken, it says the framework is overkill.
WHY NOT JUST ASK YOUR AI
Asking your AI to critique a plan gets you a critique
Ask any AI to review a plan and it will find things wrong, because that is what you asked for. The findings arrive fluent, plausible, and impossible to check.
This works differently in three ways:
- Every finding quotes your document, with its location, so the conversation is about evidence instead of taste
- The verdict is defined, not improvised. Four ratings with written definitions mean two audits of two plans can sit side by side
- It is built not to over-find. A plan that names a hard obstacle it cannot yet solve is facing the challenge, and is rated promising. When the situation is routine and nothing is broken, it says the framework is overkill instead of manufacturing a diagnosis
That last one is the difference between a test and a performance of rigor.
FAQ
Questions before you buy
Which AI tools does it work with?
Any AI tool you already use, such as ChatGPT, Claude, or Gemini. You download the files and add them to your workspace, or attach them to a conversation.
The output is markdown, so nothing else is needed. The audit works from the document you supply, so there is no connector to set up and nothing to install.
Do I need to have read Good Strategy/Bad Strategy first?
No. The files carry the method: what a diagnosis, a guiding policy, and a coherent action set are, how to tell each one from the things commonly mistaken for it, and what to do when one is missing.
If you have read the book, this is the part a book cannot be: a repeatable procedure, with extraction moves, pass tests, a report structure, and worked repairs.
Is this connected to Richard Rumelt or his publisher?
No. It is an independent application of the Good Strategy/Bad Strategy framework. There is no affiliation, endorsement, or licence.
It is not the book and does not replace it. The included evidence bank summarizes the book's cases compactly and attaches a caution to each, including where a figure is the author's own reconstruction rather than independently verified.
Isn't this just a prompt I could write myself?
A prompt gets you an opinion. This is a written procedure, and most of the value sits in the parts a prompt leaves out:
- Fifteen diagnostic lenses, each with the question to ask, the red flag, the good sign, and the fix
- A nine-step repair sequence, from naming the real challenge to stating the trade
- Four verdict definitions, so results stay comparable between plans and over time
- Eight fairness rules the audit checks itself against before it delivers
Will it just tell me my plan is bad?
No. The verdict has four settings and two of them are not failures. A plan with an intact kernel is rated good strategy. A plan whose kernel is present but has one weak element is rated promising but incomplete, and the repair is scoped to that element rather than a rebuild.
A plan that names a hard obstacle it cannot yet solve is treated as facing the challenge. If the situation is routine and nothing is broken, the audit says the framework is overkill rather than manufacturing a problem.
My plan is a slide deck, or it was never written down. Can it still be audited?
Yes. There is specific guidance for decks (read the headlines in order and test whether they form a causal chain from situation to approach to action), for single paragraphs and fragments, and for strategies that exist only in conversation or email.
If there is no written strategy at all, it can work backwards: list the policies that are genuinely unusual about how you operate, and find the single challenge they serve. If there is no hidden logic to find, it says so rather than inventing one.
How do I know it will not make things up about my business?
It is written not to. Every candidate diagnosis is labeled a hypothesis and arrives with the facts that would make it true and the one observation that would refute it. You decide which reading is right.
Where a needed fact is missing from your document, it names the fact and suggests a cheap way to get it, such as reviewing the last three attempts or cutting the data by cohort, instead of guessing.
What is in the download, and what does it not do?
Seven files: the skill itself, four references (kernel and hallmarks, the fifteen diagnostic lenses, the repair playbook, and the evidence bank), the audit report template, and a complete worked example.
It does not draft a strategy from scratch, brainstorm, run a SWOT, or do market research. It also tests whether the reasoning in a plan is present, not whether it is right: a well-argued diagnosis of the wrong problem passes the structure check. The audit says so when that happens, and prescribes a test instead of declaring the plan correct.
Ready in minutes
From download to working in minutes.
No app to learn, no software to install, no setup project. Each system arrives as ready-to-use files that work in any AI tool, so getting started takes minutes.
Download the files
Your download link is ready as soon as checkout completes. No account to create and no password to set.
Add them to the AI tool you already use
ChatGPT, Claude, Gemini, or any other AI tool. Upload a skill, attach a playbook, or open a template, the same way you add any document.
Put it to work
Start on the real job right away. The structure and standards are already built in, so there is nothing to configure first.
OPERATOR SKILL
Strategy Audit Skill
Audit any plan, deck, or brief against the kernel of good strategy, get a verdict with your own words quoted as the evidence, and get the rewritten diagnosis, policy, and actions.
- Instant access
- Lifetime updates
- Secure purchase