What the product decides so the user doesn't have to — prefilled forms, remembered choices, resume where you left off — and every question it had no right to ask.
Free & open · no signup · read-only — it ends by asking · nothing leaves your machine
You are working inside this repo. Mission: find every decision the product pushes onto the user that it could have made itself. The products that feel effortless are the ones doing quiet work up front — inferring, remembering, prefilling — so the user's first move is the real one. Every unnecessary question is a small resignation.
Read-only pass. Read the forms, settings, onboarding, and state persistence in code, and run the first-run and return paths; your only write is the report file.
Phase 2 sweeps the codebase through every one of these, citing file and line for each finding.
the first run produces value with zero setup; every gate between install and first output that a default could remove
fields requesting what the system already has (email on file, detected timezone, last-used project); the form that starts blank in a product full of context
the option picked three times becomes the default; sort orders, view modes, and last-used values persisted per user, not reset per session
reopening lands where the user left off: the open document, the half-done task, the scroll depth — not the home screen
after each core action, the probable next action is pre-staged one step away, not re-navigated from scratch
options that exist because the team wouldn't choose; every toggle is a question exported to the user, so which could a strong default retire
every inference visible and reversible in one step; a wrong guess costs a click, never a hunt — anticipation, not a trap
One structured report at the repo root — or in reports/, if you keep one — the same shape every time, ready for a teammate — or the optional Studio — to act on.
Every Goal Prompt follows the same four steps, so results are consistent and repeatable — no matter which one you run.
Collect everything the product asks.
Cite the field, setting, or flow and its file for every finding.
Rank by frequency × friction.
Create DEFAULTS.md at repo root.
Copy it and paste it into your agent inside the repo you want checked.
Install the goal plugin once — two commands — then just type /goal:smart-defaults.
/plugin marketplace add GhostlyGawd/goal-prompts/plugin install goal@goal-promptsOr install only this Goal Prompt as /goal-smart-defaults:
curl -fsSL https://goal-prompts.vercel.app/install | BRIEF=153 shLet an agent fetch it mid-conversation, or pull the raw Goal Prompt by URL.
https://goal-prompts.vercel.app/raw/153.mdNothing hidden — this is the whole Goal Prompt, verbatim. Read it in a minute, edit it, or copy it as-is.
# Goal: Smart Defaults & Anticipation Audit You are working inside this repo. Mission: find every decision the product pushes onto the user that it could have made itself. The products that feel effortless are the ones doing quiet work up front — inferring, remembering, prefilling — so the user's first move is the real one. Every unnecessary question is a small resignation. Read-only pass. Read the forms, settings, onboarding, and state persistence in code, and run the first-run and return paths; your only write is the report file. ## Phase 1 — Inventory the questions - Collect everything the product asks: onboarding steps, form fields, settings, pickers, empty dropdowns, confirmation choices. - For each, note what the product already knows at that moment (locale, prior behavior, context, sensible convention) that could answer it. - Trace what persists per user between sessions — and what resets to factory every visit. ## Phase 2 — Audit through 7 lenses Cite the field, setting, or flow and its file for every finding. 1. **Works before configured** — the first run produces value with zero setup; every gate between install and first output that a default could remove 2. **Never ask what you know** — fields requesting what the system already has (email on file, detected timezone, last-used project); the form that starts blank in a product full of context 3. **Choices remembered** — the option picked three times becomes the default; sort orders, view modes, and last-used values persisted per user, not reset per session 4. **Resume, don't restart** — reopening lands where the user left off: the open document, the half-done task, the scroll depth — not the home screen 5. **The likely next step** — after each core action, the probable next action is pre-staged one step away, not re-navigated from scratch 6. **Settings as indecision** — options that exist because the team wouldn't choose; every toggle is a question exported to the user, so which could a strong default retire 7. **Cheap to overrule** — every inference visible and reversible in one step; a wrong guess costs a click, never a hunt — anticipation, not a trap ## Phase 3 — Curate - Rank by frequency × friction: a needless question in the daily loop outranks one at setup. - Separate "remove the question" (default it), "answer it from context" (prefill), and "remember the answer" (persist). - For each, name the default you'd ship and the override path. ## Phase 4 — Report Create `DEFAULTS.md` at repo root: 1. **Question findings** — each: lens · flow · file · what's asked · the default that retires it 2. **The memory map** — what persists per user today versus what should 3. **First-run delta** — steps between install and first value now, and after these defaults 4. **Retired settings** — the options a confident default would remove Start the report with today's date. If `DEFAULTS.md` already exists from a previous run, read it first and lead with what changed since. ## Rules - Every proposed default names its evidence — the signal that makes it the right guess - A default that guesses wrong expensively (destructive, billed, sent) is worse than asking; flag those as correctly asked - No user-facing configuration or input surface in this repo? Say so in a one-paragraph null report and stop — a null result is a valid finding. - If a `reports/` directory exists at the repo root, write the report there instead of the root. - Before asking, present the top findings as a ranked list in plain words - Report only — end by asking which question the product should stop asking first
reports/ directory exists at the repo root, write the report there instead of the root.Curated neighbors — briefs that answer the adjacent question, worth running in the same session.
The first session: empty states, a guided first action, and the fastest path to a first win that leaves a newcomer inspired to build the next thing.
The screens users meet with nothing yet — first run, no results, cleared, deleted, error-empty — and whether each teaches the next step or just shows a blank and a shrug.
The hundred small behaviors between works and feels good — undo, preserved state, focus, forgiveness — the joinery a user can't name but always notices.
The hundred small behaviors between works and feels good — undo, preserved state, focus, forgiveness — the joinery a user can't name but always notices.
Not how fast it is — how fast it feels. Instant acknowledgment, optimistic writes, skeletons over spinners, honest progress for the waits that remain.
What grows the more it's used — history, saved work, personalization — and whether it's surfaced back as value or sits inert in a table nobody reads.
The moments users remember and retell — the first win, the share-worthy output, the one interaction worth a demo — placed with restraint, never confetti on command.
Runs inside these playbooks — curated sequences you can launch with one paste: