More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Probabilistic Thinking: Turn Confidence into Numbers

    1

    Turn vague confidence into numbers — anchor on base rates, express ranges instead of points, update beliefs with explicit likelihoods, and Fermi-bound the unmeasurable.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Probabilistic Thinking: Turn Confidence into Numbers

    Probabilistic Thinking: Turn Confidence into Numbers

    Example session with this skill installed

    The team says the new checkout flow will probably lift conversion. Turn that into a number we can check, update, and decide on.

    • Read your context and instructions
    • Compiled the probabilistic thinking

    Stated a checkable claim first: the new checkout flow lifts conversion within 30 days of launch — outcome, timeframe, unit — instead of 'will probably go fine', which cannot be falsified. Anchored on the base rate: past checkout redesigns at comparable shops lift conversion about 40 percent of the time, median lift when they work about 5 percent. The counter-case was written explicitly: this redesign fixes the single biggest drop-off point measured in session replays — a specific, evidenced reason to deviate upward — so the prior moved to about 55 percent chance of any lift. Ranges, not points: the 80 percent interval on the lift was minus 2 percent to plus 9 percent, widened because the outside view was thin and overconfidence was assumed. The verbal mush was translated into bands so the team could argue with the numbers instead of the adjectives. Evidence arrived a week later: plus 3 percent with wide error bars. The likelihood ratio modestly favored real lift over noise, so the posterior nudged to about 60 percent — updated in the open, with yesterday's posterior as today's prior. A Fermi check on engineering cost versus expected revenue showed that even at the low end of the range, the payoff clears the cost by 4x — so the decision to ship is stable across the whole interval, and further estimation is unnecessary. The estimation stopped there, by design. The forecast, its interval, and the date went into the estimation log, so the next monthly revisit can count whether reality landed inside the interval — and if the hit rate is 40 percent instead of 80, the intervals get widened mechanically until the hit rate matches.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Likely is not a number. Neither is probably fine or low risk. Decisions made on vibes inherit the vibes' calibration, which is to say none. This tool converts beliefs into numbers with stated uncertainty — not because the numbers are true, but because numbered beliefs can be checked, updated, and argued with. This skill covers the two traps it defeats — false precision and base-rate neglect — and the move: state a checkable claim, anchor on a base rate then argue with it, give ranges instead of points, update in the open when evidence arrives, and Fermi-bound what cannot be measured. It includes a worked pass on a checkout redesign, calibration drills — the estimation log and the base-rate habit — and the guards: if the decision is invariant across the whole range, stop estimating; label guesses as guesses; and match precision to uncertainty, because three significant figures on a 5x-uncertain product is lying with math.

    It converts beliefs into numbers; it does not make the numbers true — numbered beliefs can be checked, updated, and argued with.
    No base rate, no reference class, no counter-case means the result is a hunch, and the skill says so instead of dressing it up.
    Never Fermi what you could just look up or measure — the decomposition is for genuinely unmeasurable quantities.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    A point estimate with no range is a confession that you have not thought about uncertainty. Give at least one confidence interval — a 50 percent and an 80 percent band is a good default — and widen when the outside view is thin.

    How often this class of thing happens — past migrations, industry incident rates, historical velocity. The outside view comes before the gut, and the gut has to argue against it with evidence.

    When the decision is invariant across the whole range. The estimate served its purpose; precision beyond the decision is vanity.