MBA Crusher

An open skill suite that trains AI agents to work like a reliable first-year MBA associate. It turns vague business requests into structured, manager-ready first drafts with explicit assumptions, risk flags, and escalation paths.

Built for OpenClaw and compatible agent platforms. Tested on GPT-5.4.

What it does

Give your agent a business task — "figure out whether this new product line is worth pursuing" or "our webinar registrations are low, diagnose why" — and instead of a chatbot-style answer, you get back a structured deliverable: triage of which functions are involved, the analytical workplan, explicit facts/assumptions/estimates/unknowns, a defended recommendation, and a review block your manager can sign off on.

The acceptance criterion: a manager can use the output with less than 20% rework.

Modules

Module	What it covers
Core	Triage, assumption handling, Review Block, escalation discipline, analytics-light
Marketing	Funnel diagnosis, ICPs, campaign briefs, positioning, pricing perception
Strategy	Market entry, competitive response, growth options, disruption analysis
Finance	Margin diagnosis, unit economics, investment analysis, forecasting, valuation
Analytics	KPI design, experiment readouts, cohort/funnel analysis, metric interpretation
Product / GTM	PMF signals, feature prioritization, launch planning, business models
Sales / BD	Pipeline diagnosis, deal qualification, account analysis, partnerships
HR / People Ops	Org design, attrition diagnosis, candidate evaluation, change management
Operations	Process diagnosis, capacity planning, vendor evaluation, SOP design
Executive Comms	Pyramid principle, storylining, audience adaptation, data storytelling

Each module includes a SKILL.md (the runtime instruction set), reference libraries (deep frameworks loaded on demand), templates, and worked examples.

Architecture

Core Light ─── always loaded, governs structural discipline
    │
    ├── Marketing
    ├── Strategy
    ├── Finance
    ├── Analytics
    ├── Product / GTM
    ├── Sales / BD
    ├── HR / People Ops
    ├── Operations
    └── Executive Comms
         │
         └── references/ templates/ examples/ ─── lazy-loaded on demand

Core is thin and always available. Functional modules activate based on the task. Reference libraries load only when the analysis needs deep framework detail. This keeps context budgets tight and model compliance high.

Structural guarantees

Every substantive output includes these structural elements, enforced by compliance rules in every module:

• Triage Block — identifies which business functions are involved when a task is cross-functional
• FAEU sections — literal Facts, Assumptions, Estimates, Unknowns labels that make the evidentiary basis visible
• Defended recommendation — a default recommendation with explicit override conditions
• Review Block — a 6-field structured sign-off (decision needed, key assumptions, confidence level, risks, escalation needed, next action)
• No conversational offers — the agent produces deliverables, not "If you want, I can..." prompts

Benchmark results

Tested across 30 benchmark tasks spanning all 10 modules. Scored on 6 binary structural checks per task.

Metric	Result
Full pass rate	9/10 sampled tasks
Review Block compliance (correct 6-field format)	100%
FAEU compliance	100% (Round 3)
Conversational offer pattern	0% (eliminated)
Hard failures	0

Benchmark tasks, scoring rubric, and gold-standard outputs are included in benchmark-set/ so you can run your own evaluation.

Quick start

OpenClaw

1. Clone this repo
2. Copy the skills/ folder into your OpenClaw workspace
3. In your agent configuration, load the skills you need:
   • Always load mba-associate-core-light
   • Add 2-3 functional modules based on your use case
4. Send a business task and review the structured output

Recommended loading profiles

General business assistant: Core + Marketing + Strategy + Finance

Product team: Core + Product/GTM + Analytics + Strategy

Sales/revenue team: Core + Sales/BD + Finance + Marketing

People/operations: Core + HR/People Ops + Operations

Full suite: Load all 10 modules (works well on GPT-5.4-class models)

Repository structure

mba-associate-suite/
├── README.md
├── LICENSE
├── AGENTS.md
├── .gitignore
├── docs/
│   ├── architecture.md
│   ├── product-vision-v2-1.md
│   ├── build-plan.md
│   ├── skill-authoring-guide.md
│   ├── deployment-profiles.md
│   ├── runtime-efficiency-principles.md
│   ├── packaging-and-pricing.md
│   └── ...
├── skills/
│   ├── mba-associate-core-light/
│   │   ├── SKILL.md
│   │   ├── references/
│   │   ├── templates/
│   │   └── examples/
│   ├── mba-associate-marketing/
│   ├── mba-associate-strategy/
│   ├── mba-associate-finance/
│   ├── mba-associate-analytics/
│   ├── mba-associate-product-gtm/
│   ├── mba-associate-sales-bd/
│   ├── mba-associate-hr-people-ops/
│   ├── mba-associate-operations/
│   └── mba-associate-exec-comms/
└── benchmark-set/
    ├── tasks.md
    ├── tasks-phase-2-expansion.md
    ├── tasks-phase-3-exec-comms.md
    ├── scoring-rubric.md
    ├── evaluation-protocol.md
    └── gold-standard-outputs/

What this is not

This is not a CEO replacement, a framework encyclopedia, or a generic productivity bot. It is a disciplined junior associate layer that prepares structured first-pass work so a manager can review, correct, and approve it quickly. It explicitly flags what it doesn't know and escalates what it shouldn't decide.

License

Deft Source Available License — based on Apache 2.0 with an additional commercial restriction. Free for internal use, education, and research. For-profit entities may not resell or bundle this in paid services without explicit permission. See LICENSE for details.

Contact

Maneek Mohan — maneek.mohan21@gmail.com