Plugin Architecture and Installation
In Lesson 1, you learned that the DLA Stack is a three-methodology innovation system and that AI compresses the execution overhead at every stage. This lesson installs the toolkit: the Cowork innovation plugin with 10 commands covering every stage of the DLA Stack, plus the innov.local.md configuration file that makes every output specific to your venture.
Think of this as fitting out your workshop before starting to build. By the end of this lesson, you will have all 10 commands available in Cowork and understand what each one does — and you will have the innov.local.md template ready to fill in as you work through the chapter.
Plugin Architecture
The innovation plugin lives in the agentfactory-business-plugins repository. When you install it, you get the skills and configuration — not the development artifacts. Here is what you install:
.claude-plugin/
└── plugin.json ← Plugin metadata and configuration
skills/
├── idea/SKILL.md ← /idea — idea generation + evaluation
├── discovery/SKILL.md ← /discovery — customer research + synthesis
├── hypothesis/SKILL.md ← /hypothesis — assumption mapping + MVP design
├── canvas/SKILL.md ← /canvas — Business Model Canvas
├── financials/SKILL.md ← /financials — unit economics + runway model
├── pitch/SKILL.md ← /pitch — investor deck narrative
├── sprint/SKILL.md ← /sprint — innovation sprint planning
├── market/SKILL.md ← /market — competitive intelligence + sizing
├── gtm/SKILL.md ← /gtm — go-to-market strategy
└── validate/SKILL.md ← /validate — build-measure-learn analysis
agents/
├── idea-generator.md
├── customer-intelligence.md
├── business-model-architect.md
└── fundraising-readiness.md
README.md
LICENSE
The plugin repository also contains development artifacts — evals/, tests/, examples/, specs/ — used by the plugin maintainers. You install only what is above: the plugin configuration, skills, and agents. The development artifacts are for contributors, not learners.
The 10-Command Map
Each command handles a specific part of the DLA Stack. Here is the full reference:
| Command | DLA Stage | Primary Use |
|---|---|---|
/discovery | Design Thinking | Customer interview guides, JTBD synthesis, pain ranking, HMW problem statements |
/idea | Design Thinking | Idea generation (100 ideas), DVF scoring, pressure testing, shortlisting |
/hypothesis | Lean Startup | Assumption mapping with risk scores, MVP scoping, validation test design |
/validate | Lean Startup | Build-Measure-Learn analysis, pivot-or-persevere decisions |
/canvas | Lean Startup + Cross-cutting | Business Model Canvas build, stress-testing, alternative model exploration |
/financials | Cross-cutting | Unit economics, 18-month runway model, scenario analysis, Series A readiness |
/pitch | Cross-cutting | Investor deck narrative, Q&A preparation, executive summary |
/sprint | Agile | Innovation sprint planning, user story writing, retrospective with assumption updates |
/market | Cross-cutting | Competitive intelligence, market sizing (bottom-up), differentiation map |
/gtm | Cross-cutting | ICP definition, channel strategy, pricing model, 90-day GTM calendar |
The DLA flow looks like this:
DESIGN THINKING ──► LEAN STARTUP ──► AGILE
/discovery /hypothesis /sprint
/idea /validate
/canvas
↓ applies at any stage:
/market /financials /pitch /gtm
Stage-Aware Calibration
Every command checks your venture stage and warns if you are at the wrong DLA level for what you are asking. This is not a restriction — it is a guardrail. Examples:
/financialsat IDEA stage: "You do not have customer data yet. These numbers will be highly speculative. Do you want to proceed with scenario modelling, or return here after customer discovery?"/ideaat MVP stage: "You have a validated product. Are you pivoting or exploring adjacent markets?"/sprintbefore assumptions are validated: "Your assumption stack has 8 untested HIGH-risk assumptions. Running sprints before validating these risks building the wrong thing."
These warnings surface the failure modes from Lesson 1 in real time.
Installation
Prerequisites: You need a Cowork account and access to the agentfactory-business-plugins repository. If you have not used Cowork before, complete Chapter 28 (Productivity Domain Agents) first — this plugin works alongside work.local.md from that chapter.
Step 1 — Open the plugin browser. In Cowork, click the sidebar gear icon → Customize → Browse plugins.
Step 2 — Add the marketplace. Under Personal, click + → Add marketplace from GitHub → enter:
https://github.com/panaversity/agentfactory-business-plugins
Step 3 — Find and install the Innovation plugin. Browse the catalog, find Innovation, and click Install.
Step 4 — Verify installation. In a new Cowork session, type:
Tell me about the innovation plugin. What commands are available
and what does each one do?
You should receive a response listing all 10 commands with descriptions. If Cowork does not recognise the plugin, close and reopen the session — plugin activation sometimes requires a fresh session.
The innov.local.md Template
innov.local.md is the configuration file that makes every innovation command output specific to your venture rather than generic. Without it, /canvas produces a generic Business Model Canvas. With it, /canvas builds a canvas using your actual customers, your validated assumptions, and your current traction data.
The template has eight sections:
| Section | What It Configures | When to Fill In |
|---|---|---|
venture: | Name, stage, problem statement, target customer, solution hypothesis, unfair advantage | Now — even rough is better than blank |
key_assumptions: | Your assumption stack with risk levels and test status | Lesson 5 (Assumption Stack) |
customer_profiles: | Personas from discovery — JTBD, pains, buying process | Lesson 3 (Customer Discovery) |
business_model_canvas: | Canvas blocks with evidence quality | Lesson 8 (Business Model Canvas) |
financial_model: | Unit economics, runway, milestones | Lesson 9 (Unit Economics) |
competitive_landscape: | Competitors, alternatives, differentiation, moat | Lesson 10 (Market Intelligence) |
fundraising: | Round, pipeline, data room status | Lesson 12 (Investor Pitch) |
intrapreneurship: | Internal approval pathway, budget, stakeholders | As applicable |
Right now: Copy the template and rename it innov.local.md. You do not fill it in fully until the Lesson 15 capstone. But if you have a venture idea already, fill in the venture: section — a rough problem statement and target customer is enough to improve every command output immediately.
The Validation Test
At the end of Lesson 15, you will run four validation prompts to confirm your innov.local.md is producing specific rather than generic output:
/idea — what 3 ideas should I explore this week?
Expected: Ideas specific to your problem space, not generic innovation ideas.
/canvas — what is the health of my business model canvas?
Expected: Block-by-block assessment using your actual canvas content.
/hypothesis — what is my most critical untested assumption?
Expected: A specific assumption ID with a cheap, concrete test recommendation.
/pitch — write my executive summary
Expected: A summary with your actual traction numbers and problem statement.
If the outputs are generic, a section needs more specificity. The most common gaps: customer_profiles (pains too vague), key_assumptions (too few), competitive_landscape (differentiation too generic), financial_model (all ASSUMED).
The innov.local.md template includes an intrapreneurship: section specifically for you. Use it to record your approval pathway (current stage, next gate, gate owner, gate criteria), internal constraints (budget, headcount, technology), and stakeholder map (champion, neutral, resistant). This replaces the fundraising: section if you are not raising external capital — though you can use both if your intrapreneurship project eventually leads to a spin-out.
Mini-Activity: Configure Your Venture Section
Before moving to Lesson 3, do one thing:
Copy the template, rename it innov.local.md, and fill in the venture: section.
If you have a venture idea in mind:
/idea
I am starting a new venture. Here is my rough concept:
[Describe your idea in 2-3 sentences — the problem, the customer,
and your initial hypothesis for the solution]
Help me fill in the venture: section of innov.local.md with this
information. Ask me clarifying questions if my description is too
vague to complete any field.
If you do not have a venture idea yet, use the AP automation example:
I am following the AP automation worked example from Chapter 40.
The venture is a B2B SaaS for accounts payable automation targeting
CFOs of mid-market companies ($5M-$50M revenue). The core problem:
manual invoice reconciliation, WhatsApp approvals with no audit
trail, no real-time AP visibility.
Fill in the venture: section of innov.local.md for this example
so I can follow along with the chapter exercises.
This gives every command real context to work with from the start.
Try With AI
Use these prompts in Cowork or your preferred AI assistant.
Reproduce — List all 10 commands:
List all 10 innovation plugin commands. For each one, describe:
1. What DLA stage it operates at
2. What input it needs
3. What output it produces
4. One example of when you would use it
What you are learning: Understanding what each command produces before you use it means you know what to expect and how to evaluate the output. You are not discovering the tool as you go — you are using it with intent.
Adapt — Stage mapping for your context:
I am at the [IDEA / DISCOVERY / VALIDATION / MVP] stage of my
innovation project. Which innovation plugin commands should I
use most at this stage, and in what order? Which should I avoid
until later, and why?
What you are learning: Stage-aware command selection is what separates a systematic innovation process from random tool use. This prompt forces you to think about where you are before deciding what to do next.
Apply — Set up your venture context:
I am [an entrepreneur / an intrapreneur at {company}] working on
[describe your idea or project]. Help me fill in the venture:
section of innov.local.md. Ask me clarifying questions if any
field is unclear. After we complete it, run /idea to test whether
the output is specific to my venture.
What you are learning: The difference between generic and specific AI output is almost entirely determined by the quality of context you provide. A well-filled innov.local.md is the difference between "here are some business ideas" and "here are three ideas specific to your CFO customer's biggest pain point."
Flashcards Study Aid
Continue to Lesson 3: Customer Discovery and Problem Statement →