Backbrief
Backbrief, the base. Grade the idea. Get the plan. Before the work costs money.
A council of advisors for Claude Code. It writes your business plan from your own numbers and grades it before the work costs money.
- A business plan written from your own numbers
- A scorecard that names the plan's weaknesses and says what to do next
- The next ninety days as units, each with an owner and a pass or fail test
$29, paid once. One license, one person.
Runs inside Claude Code, Anthropic’s coding assistant. You need it installed and signed in, and a project folder.
All sales are final: refund policy and license terms
A real grade
What a scorecard looks like
Backbrief graded a video that pitches a money-making method: flipping Amazon return pallets. Watch it graded.
Graded on Backbrief 0.1.1 on 2026-10-06, before verdicts were added in 0.3.0. A scorecard made today also carries a verdict on what to do next and, when a gap is one only evidence can close, the cheapest test that would settle it.
The idea
Buy manifested pallets of Amazon customer returns from a liquidation warehouse or an auction site when the eBay sold value of the top five items is triple the pallet price, pick the pallet up by truck, test the gear, and resell it on Facebook Marketplace the same day; "people are clearing $5,000 to $10,000 a month".
Source: A public Facebook reel by Decode Power.
| Clarity of offer | 5.6 | |
|---|---|---|
| Market realism | 3.7 | |
| Financial viability | 2.6 | |
| Execution feasibility | 3.6 | |
| Risk coverage | 3.3 | |
| Differentiation | 2.5 | |
| 8.5 pass line | ||
Average 3.6/10. Grade C.
Lowest: Differentiation (2.5). The reel's only edge is a screening rule any bidder can run on the same public lots, and "being the middleman for Amazon's junk" is a commodity role; the free warehouse map shows the method is distributed, not owned.
What it is not
Backbrief does not run your business. Publishing, sending and spending stay with you. It gives analysis, not advice. A grade is a stress-test: it predicts nothing and certifies nothing, and it names the weaknesses you would otherwise pay to find.
From the idea to the plan
- The brief
/intaketurns what you paste in into a business brief. It asks only for the essentials it cannot find, and it marks every estimate as an estimate.- The plan
/business-planwrites the plan from the brief: the offer in one sentence a stranger understands, the market evidence, the model with the costs recomputed from your own rates, the channels ranked by what each costs you, the top risks (found by a pre-mortem that assumes the business failed six months in) with a mitigation or your acceptance of each, the reason to pick this over the obvious alternative, written as an answer to that alternative’s strongest case, and the next ninety days as units with an owner, a dependency and a pass or fail test on each. Unknowns stay unknown, and the grade prices them.- The grade
/gradesends the plan to the advisors, each in a fresh context with the plan and the rubric, so no advisor sees another’s scores. The scorecard says how proven the plan is and what to do next: commit, test next, fix the plan, walk away, or stop at the cap. When a gap is one only evidence can close, it names the cheapest test that would settle it. The revise loop has a hard cap and does not spend passes rewording a gap only evidence can close unless you ask.- The go
/approverecords your decision. On a go, the plan’s units route to Backbrief’s own agents, researcher, builder and runner, with the verifier gate on each. A test go runs only the scorecard’s next test, inside its caps, at any grade. Publishing, sending and spending stay yours.
Other doors in
/grade-idea- A raw idea in a sentence or two, no plan document. The council scores it in one pass. If it holds up,
/intakestarts the loop above. /grade-video- Paste a link to a video that pitches a money-making method. It pulls the transcript, tags every claim, and grades the opportunity.
/pick-niche- For the person with no idea yet. It interviews you on skills, hours, money and limits, then returns graded candidates, each with a first move for this week. You choose.
Want the commands in order, from install to your first grade? How to use Backbrief.
How the grade works
The rubric is public. Each score runs 1 to 10 and cites the plan's own words. No citation, no score.
- Clarity of offer.
- Market realism.
- Financial viability.
- Execution feasibility.
- Risk coverage.
- Differentiation.
Each advisor runs as its own agent in a fresh context, with the plan and the rubric, so no advisor sees another's scores. On a regrade pass, an advisor also receives its own earlier scores. The rubric, the tags and the cap are rule files in your project.
The advisors score through distinct lenses.
- The contrarian argues why it fails.
- The first-principles advisor rebuilds it from base truths.
- The expansionist argues the bigger option you missed.
- The outsider knows nothing about the industry and asks the common-sense questions.
- The executor says what you do Monday.
The scorecard reports where they disagree.
Every figure that moves a score gets a tag. Verified means checked against a named source. Unverified means plausible and unchecked. Vendor claim means the source sells something. In a video grade, a number from someone selling something never lifts the score above what an unverified number would.
Every scorecard also carries a verdict (commit, test next, fix the plan, walk away, or stop at the cap) and, when a gap is one only evidence can close, the cheapest test that would settle it. The revise loop caps at three grading passes per evidence cycle, and a new cycle starts only from a test result you confirm. Below A- at the cap, it stops and hands you the scorecard, the top unresolved weaknesses, the next test if one is named, and your options: run the test, proceed at the real grade, pivot or kill. It does not auto-proceed. An A- means no confident weaknesses are left unnamed. It does not mean the business will succeed.
The proof
The loop above wrote the plan in this example before it graded it, and the builder found the first defect while writing the plan. Backbrief ships with its own example: TrailNotes, an invented newsletter business. Its figures are the example's premise. The grading panels that ran on it were real. It went through /intake, /business-plan and /grade to the cap. The average moved from 7.1 to 7.4 to 7.6, a B every pass, short of the 8.5 pass line. Backbrief did not round a B up to a pass. The example ships in the download, so you can read a hard grade before you run your own.
| Grading pass | Average on a scale to 10 | Score | Grade |
|---|---|---|---|
| First pass | 7.1 | B | |
| Second pass | 7.4 | B | |
| Final pass | 7.6 | B | |
| 8.5 pass line | |||
Backbrief's own builder agent, loading the unit-economics skill, recomputed the example's costs from the rates in its brief. They summed to $391.76 a month against the brief's "about $340/month, all-in":
a $51.76 gap, 15.2% of the stated figure, that cannot be closed from the brief's numbers.
Then:
The plan does not pick between the two totals.
A small set of real YouTube pitch videos went through /grade-video too. They graded C, with an average below 7.0, and each grade comes with a ledger that tags the claims behind it. The grade covers what the video alleges. It does not accuse the presenter, and an unverified claim is unchecked, not false.
See the full scorecard
The score of each advisor, the panel’s read and the first move it named for the Amazon return pallets pitch. It opens on this page.
What you need
Claude Code, installed and signed in. A project folder. No other accounts, no keys.
Backbrief runs on your own Claude Code access, whatever plan or billing it uses, and a grade uses that access like any session. The product has no server and collects nothing; this website’s own collection is on the privacy page. Video grading can use free local tools (yt-dlp, ffmpeg), and Claude asks before installing any.
Who it is for
For you if you are about to spend money, time or reputation on a business idea, a plan or a video pitch, and you want the plan written and graded before you commit. You work in Claude Code or will install it for this. It also fits you with no idea yet.
Not for you if you want someone to run the business for you. Backbrief writes the plan and grades it. After your go, its agents work the plan’s internal units, and you keep the outward actions: publishing, sending and spending.
Pricing
$29, paid once.
One license, one person. Updates are free for as long as Backbrief is offered. Backbrief is the base of a line: Site Check, Claim Check and Deal Check are separate products, each with its own license. All sales are final; read the license and the refund policy before you buy.
Other Backbrief products
Each one is bought on its own, and none needs another first.
- Site Check
You are about to launch a website, or you want to know why AI assistants do not cite yours.
$79, paid once.
- Claim Check
You are about to send an ad, an email or a post.
$79, paid once.
- Deal Check
You are about to sign or pay for a proposal, quote or contract.
$79, paid once.
FAQ
Does it tell me whether to do it?
It tells you what to do next, and why. For a plan the verdict is commit, test next, fix the plan, walk away, or stop at the cap; for a raw idea it is test next, add detail, or walk away. When a gap is one only evidence can close, it names the cheapest test that would settle it. The decision stays yours, and Backbrief does not tell you to buy what a video sells.
My idea gets a C. Now what?
Read the verdict first. On an untested idea a C is the expected result, not a verdict against the idea. For a raw idea, /grade-idea says test next, add detail, or walk away and sketches the cheapest test; adding detail and grading again is free. For a full plan, /grade can also say fix the plan, and you can approve its next test on its own before committing to the plan, then regrade with the result. A walk away verdict names the evidence behind it. The decision stays yours.
Can I use it for client work?
Yes. The plans, grades and reports you produce are yours to deliver. The product is licensed to one person, so a client who wants it in their own project needs their own license.
What does it fetch from YouTube?
The transcript, and only when you say go ahead. Before any fetch, /grade-video tells you the platform's terms may restrict automated download and offers a local file or a transcript you have the right to use first. It never logs in to a platform.
Is this legal or financial advice?
No. Nothing Backbrief produces is legal, financial, tax or investment advice. A claim about local law stays an unverified assumption until it is researched, and licensed professionals answer those questions.
How long does install take?
About two minutes. Unzip the download, open Claude Code in the folder and type: read the README.md in this folder and install this for me. Claude copies the pieces and runs /verify-install.
What do I get at the end?
A business brief in your own words, a business plan built from it with the next ninety days as units you can hand to a person or an agent, a scorecard that cites the plan's own sentences with a verdict on what to do next, and your recorded decision. All of it is plain Markdown in your project folder.
Not ready to buy? See the full sample scorecard.
Grade the idea. Get the plan.
$29, paid once.
All sales are final: refund policy and license terms