I use techniques like these in my own client work — opportunity recognition, business case building, AI strategy, and more. This is the open part of that toolkit: install the pieces you need, and use each one the way you'd use a trusted colleague's playbook, not a black box.
public-sector-ai-service-design
(a public-sector lens on top of the core AI-strategy and business-case
skills: opportunity screening for public value, stakeholder/political
landscape mapping, procurement navigation, and a six-element
decision-readiness model for public decision bodies) — plus 5 new
skills spread across existing packs, each checked against a
10-capability practitioner report before being added, so the pack
grows only where there was an actual, checked gap: predicting a
specific audience's reaction to a concept
(taste-emulation-heuristic), protecting your own
judgment when working with AI
(self-efficacy-and-cognitive-rot-shielding), stripping
AI-generated jargon before it ships
(whiteboard-clarity-and-jargon-stripping), closing the
information gap with a decision-maker before a pitch
(stakeholder-pressure-and-information-gap-mapping), and
a feature-freeze discipline for AI-native products
(tiny-core-identification-and-feature-freeze). Full
detail, including what was deliberately left out and why, in the
changelog.
Business and service designers who aren't yet fully at home in demanding, technical AI environments. It's a running start if you're early in an AI business design career, or moving into AI from a different background. Already senior in this field? You probably have something similar in your own toolkit already — you'll recognize the shape of it.
Designing business models, products, and processes where AI plays a real role — from AI-native ideas built from scratch, to updating how things already work for the AI era. Think of this pack as the AI-specific layer on top of the business and service design frameworks you already know.
Six situations where reaching for this pack — instead of prompting from scratch — actually pays off. Each chains two or three packs together.
You've got a handful of AI ideas and no reliable way yet to tell which deserve real investment. Scan and size the opportunity, test whether it's a genuine value-chain reshuffle or just automation, score it against a 5-dimension model, then build the business case for the top pick — a scored shortlist, not a gut call.
An AI feature is about to go from prototype to production, and "a human checks it" isn't an actual design yet. Classify the process against a four-level HITL maturity model, build a confidence-score routing table with a named calibration owner, and specify the AI's behavior and guardrails — a concrete oversight architecture, not an aspirational policy line.
You're running — or reviewing — a client's Business Model Canvas and need to know if it's actually sound, not just filled in. Run the session, diagnose the finished canvas against seven known failure patterns, and match it against a 159-pattern innovation library — a canvas stress-tested, not just consultant intuition.
You need a credible AI prototype fast, and a demo that proves one specific point without overpromising production-readiness. Build the narrowest prototype that proves the hypothesis, frame it honestly for the audience, deliver it, then bridge straight into the ROI conversation.
You have an AI-native product idea and need to go from customer insight to something a build agent can actually act on. Go from JTBD to a scored MVP — with a feature-freeze discipline before scope creeps — to a PRD, with the human-oversight layer designed in from the start, not bolted on afterward.
You're scoping an AI idea for a city, agency, or non-profit, where "who's the sponsor" isn't one clean answer and ROI isn't the whole case. Screen the idea for public value and mandate fit, map the stakeholder types and veto points, navigate procurement, and build a decision-readiness case a public board can actually approve.
Everything here is organized into packs — folders of related skills you install separately, only when you need them. Inside a pack, a skill is one named technique: a structured way of doing one specific piece of business-design work — sizing an opportunity, building a business case, scoping an AI pilot — that you invoke by name instead of writing a prompt from scratch every time. Four packs also include a read-only agent: an optional second pass that checks a skill's output before you act on it.
Four things hold every skill in this pack to the same discipline, regardless of topic.
Each skill cites a named framework (Porter, Kirzner, Liedtka, BABOK, Minto, and others) or one of my own client engagements — nothing here is invented from scratch.
Every skill is real, usable methodology the moment it ships — I privately track how much of my own field experience is layered onto each one, as a backlog for where I deepen it next, not as a public completeness score.
Every skill says what it deliberately does not do. Unverifiable numbers are marked [assumption — verify], never presented as fact.
Every output is a decision-support draft, not a decision. You review it, decide, and carry the accountability.
Nine core packs cover situations that come up across most business-design work; four specialisation packs go deeper into one specific one. This is a directory, not the full picture — for what each pack actually does, skill by skill, see the full breakdown in the GitHub README, or click straight through to any pack below.
| Pack | Skills |
|---|---|
research-commercialisation |
12 |
ai-native-startup-design |
9 |
business-model-canvas |
19 |
public-sector-ai-service-designNEW |
7 |
Note: the specialisation packs published here are a subset — get in touch (see About this project) if you're interested in the fuller versions.
Four packs include an optional, read-only agent you can invoke separately to stress-test a skill's output before it goes to a decision-maker. None of them edit anything — each just returns a findings table for you to act on.
| Agent | Pack | What it checks |
|---|---|---|
assumption-stress-tester |
business-case-and-analysis |
Adversarially challenges a business case's assumptions before the number goes to leadership. |
market-sizing-cross-validator |
opportunity-recognition |
Cross-checks a TAM/SAM/SOM calculation with an independent top-down/bottom-up method. |
competitive-blind-spot-scanner |
business-design-frameworks |
Looks for un-scanned competitors or angles in a competitive/positioning analysis. |
ai-initiative-readiness-auditor |
ai-strategy-and-governance |
Audits an AI initiative's scoring and governance checklist for gaps before approval. |
Nothing to download by hand — no ZIP, no git clone. You
point Claude Code or Cowork at the GitHub repo below and it fetches
what it needs on its own:
github.com/Pilot2Service/AI-Business-Designer
/plugin marketplace add Pilot2Service/AI-Business-Designer
/plugin
The first command registers this repository's catalog
(ai-business-designer-skills) by reading
marketplace.json straight from GitHub — nothing is
cloned to your machine at this step. The second command opens the
plugin manager: choose Browse and install plugins,
pick the marketplace, then install the specific packs you want. Only
the packs you actually install get pulled onto your machine.
Open Cowork, look for a Customize or
Plugins area (exact wording can shift between app
versions), choose Add marketplace, and paste
Pilot2Service/AI-Business-Designer or the GitHub link
above. From there, browse and install packs the same way. If that
menu isn't where you expect, check Cowork's own settings/help for
"plugin marketplace" — this is the one part of this page I can't
fully verify against Anthropic's published docs, since Cowork's UI
isn't documented there in the same detail as the CLI.
Don't have Claude Code or Cowork yet? Set one of those up first (see docs.claude.com), then come back here. Full walkthrough, including what a good first run looks like: QUICKSTART.md.
Finding and shaping opportunities for AI-driven products and services is what this pack is actually about. As of August 2026 there are plenty of skills packs built around general service-design methodology, but relatively few aimed specifically at business design work in an AI context — that's the gap this fills.
It's built from my own experience and from the general models I think AI business work actually needs. I've spent over 15 years in business commercialization, productization, and service development, and this collection grew directly out of that work.
This is a personal project, not a commercial product. I keep updating and refining it as my own work evolves, but there's no fixed roadmap or release schedule behind the public version — it isn't exhaustive either. It contains the heuristics and decision models I actually reach for in client engagements, not a complete map of the field.
Think of it as an open-source share of part of my own toolkit — these are close to the techniques I use myself, published here for anyone to use, adapt, or learn from. It's also, honestly, a way to show what I can actually do, and how I see this context-engineering discipline developing.