review-business-plan

Report-only board review of a business plan that scores seven lenses and names the one finding that most changes it, called by ceh-business-plan:develop-business-plan.

SKILL.md

GitHub

Review business plan

Judge a business plan on the seven things experienced operators check first, and return a scored report with one headline finding. Done is the report in the Output format with all seven lenses scored, each score backed by a quote or "nothing found", and the next skill named.

Procedure

  1. Read the whole plan. Default to .agents_workspace/business/BUSINESS_PLAN.md (schema in ${CLAUDE_PLUGIN_ROOT}/references/business-plan-schema.md). Any other plan, deck, or memo the user supplies is reviewed the same way.
  2. Score each lens 0, 1, or 2 using the table below. Quote the sentence in the plan that earns the score. A lens with nothing to quote scores 0.
  3. Answer the fresh-owner question: if new owners took this plan over tomorrow with no attachment to it, what would they stop, and what would they do first? (Andy Grove asked this of Intel's memory business and exited it.)
  4. Pick the headline finding: the single lens whose failure makes the other scores irrelevant. Order of precedence when several score 0: customer, then survival, then money, then edge, reach, focus, execution. A plan nobody wants needs no cost model.
  5. Set the verdict from the scores, not from impression:
    • do not build: Customer scores 0 and the plan offers no path to a test.
    • build: no lens scores 0 and the total is 10 or more.
    • fix first: everything else.
  6. Write the report in the Output format and stop. Do not edit the plan.

The seven lenses

Lens The question Scores 2 when
Customer Who is the customer, what do they value, and what do they do today instead? (Drucker's three questions.) A named person or account, their words quoted, and their current workaround with its cost
Edge Why does this customer pick you, and what stops a better-funded rival doing the same? (Buffett's moat.) The advantage costs a rival something real to copy, and the plan says what
Focus What is the plan refusing to do? (Jobs cut Apple's line to four products in 1997.) One segment, one product, one channel, and an explicit list of attractive things declined
Money Does one unit make money, and does cash arrive before it runs out? Unit arithmetic is shown, the cash low point is dated, and the plan survives its own bad case
Reach How do the first ten customers hear about this, and what does each one cost to win? The first ten are named or listed, one channel is chosen, and a cost per customer is estimated
Survival What kills this, how would you know early, and what is the most that can be lost? (Munger: invert.) Top risks have a warning signal, a kill criterion set in advance, and the maximum loss is a number
Execution What happens in the next 90 days, who owns it, and which weekly number shows it working? (Grove's objectives and key results.) At most three objectives, each with an owner, a date, and a controllable weekly input

Score 1 is the middle case on every lens: the plan asserts an answer but offers no evidence and no test.

Rules

  • Score what is written, not what the author probably meant. A reviewer who fills gaps with charity hides exactly the holes the review exists to find.
  • Evidence is a quote, a transaction, a signed letter, a measured number, or a comparable spend. A market-size report and "everyone I asked liked it" are not evidence.
  • Treat the plan's own confidence tags as claims to check. An [evidence] tag with nothing behind it scores as an assertion, and the report says so.
  • Say "do not build this" when the customer lens scores 0 and the plan offers no path to a test. That verdict is the most valuable one a review can give.
  • Praise only what is specific. Name the strongest sentence in the plan and why it is strong, so the author keeps it through the next revision.
  • No rewrite suggestions longer than one line per lens. The fix belongs to the skill the report routes to.

Output

The review fills assets/business-plan-review.template.md, per references/business-plan-review-schema.md. Save the review to .agents_workspace/business-plan-reviews/<YYYY-MM-DD>.md, never into the plan, which this skill only reads, so the scores survive into the session that runs the next skill. Name the path in the reply.

Stop conditions

  • No plan exists, only an idea → say there is nothing to review yet and name ceh-business-plan:find-product-market-fit as the place to start.
  • The plan describes a different product than the app plan or repository it sits beside → report the mismatch as the headline finding. Every other score is unreliable until it is settled.

Hands off to

Each handoff is conditional on the lens scoring below 2, so the user chooses which to run.

  • Customer → ceh-business-plan:find-product-market-fit
  • Edge or Focus → ceh-business-plan:sharpen-strategy
  • Money → ceh-business-plan:stress-test-unit-economics
  • Reach → ceh-business-plan:plan-go-to-market
  • Survival → ceh-business-plan:run-premortem
  • Execution → ceh-business-plan:set-operating-plan