feat(setup): put rc setup on the guided rail - #144
Conversation
DX-999 Add support for paywall image localizations
Currently image localizations cannot be defined in the paywall editor. SDKs support these when viewing a paywall in the mobile app, but they cannot be defined. |
91b6c36 to
dc9f661
Compare
|
bugbot run |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
|
bugbot run |
1 similar comment
|
bugbot run |
|
bugbot run |
1 similar comment
|
bugbot run |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit a76bcf7. Configure here.
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit a76bcf7. Configure here.
Refit the account/onboarding flow (rc setup) onto the same guided rail as the store-credential setups: intro cap, step headers, rail-native selects/confirms, receipts, a spinner ledger for skills+MCP, and an outro before the agent takes over. Reuses the existing Flow/Ledger/prompt primitives. Edge cases preserved: the copyable agent prompt prints raw to stdout (no gutter); delegated sub-commands (browser login, signup, project picker) run in their own UI and the rail resumes after; MCP config returns its status/hint so the ledger shows it and hints stay on the rail; --yes still skips the launch gate; --quiet and --no-color flow through. Non-interactive/--json output is unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Keep configureAgentMCP's own doc (mcpResult no longer displaces it). MCP config is narrated (success/left-alone/manual) instead of a ledger ✓/✗ that would over- or under-state its fuzzy outcome. Declining the launch gate returns cancelled (non-zero, matching the sibling flows and the old confirmOrAbort) rather than exiting 0. Skip the one-option agent picker when no agents are installed, and give a heads-up before the browser/signup/project-picker interludes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Surface the current account (name <email>) as a receipt in the intro block when already authenticated, so rc setup confirms who you're logged in as instead of silently continuing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…logue Sign-in is a prerequisite, not a rail step. Run login/signup before the guided rail opens so another command's output never tears the rail apart mid-flow — the rail then starts clean with the account shown. When setup drives signup, suppress signup's standalone next-steps epilogue (copy-this-prompt / rc skills install / projects create), which duplicated and contradicted what setup does on the rail. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… header Revert moving auth ahead of the rail; that dropped the intro/project/location context for logged-out users. Header shows first, then Sign in on the rail. Keeps the signup epilogue suppression. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collect signup (email, name, password, keychain, terms) with the rail's own Input/Password/Confirm prompts instead of delegating to rc auth signup, and run account creation inside a ledger step with signupWithOAuth's standalone chatter suppressed. The whole sign-up now stays on the gutter instead of tearing the rail open. Reuses the shared password generator and validator. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…rompts) Add a scriptable-input seam to the rail prompts (mirrors huh's WithScriptedIO) and tests for the Flow: a gutter invariant (guards the 'rail disappears' class), quiet suppression, plain-mode glyph drop, URL no-color, and scripted Select/Confirm/Input driving real values + cancel returning ErrPromptCancelled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Autonomy (--autonomy, default auto) and skills scope (--skills-scope, default global) are now defaults shown as receipts, not prompts — cutting two questions from every run. Marketing opt-in stays a prompt (revisit when the FE handles it). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a state-aware roadmap line (sign in → pick a project → pick an agent → launch) so the user sees what's coming; drops 'sign in' when already logged in and ends at 'copy the prompt' when no agent is installed. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…setup setup surfaces this on the rail itself; keep signupWithOAuth from printing it off-gutter when it drives signup. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…on Keychain failure Bugbot findings on the signup rail: - Terms of Service confirm defaulted to Yes, so pressing Enter accepted the Terms. Default to No; acceptance now requires an explicit Yes. - A generated password could be lost silently: if the user chose to save it but the Keychain write failed, signup emitted an off-rail warning with no recovery hint, and the rail hint only fired when the user declined saving. signupWithOAuth now returns whether the password was saved; setup shows the Forgot-password recovery hint whenever a generated password ends up unsaved (declined or Keychain failure), and the off-rail Keychain warning is suppressed under setup so the rail stays intact. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Gating the Keychain-failure warning behind !fromSetup left the user-provided password case silent: setup only warned for generated passwords. Add a rail warning when a save was requested but didn't happen, regardless of whether the password was generated. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
a76bcf7 to
9782076
Compare
A mistyped confirmation returned an error and discarded the whole setupSignup flow (email, name, marketing, password choice), forcing a restart. Loop the password + confirm prompts until they match, like the old signup form did. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
bugbot run |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 2fed12a. Configure here.
The default install set now includes revenuecat-cli (the other skills reference it for CLI discovery and conventions, so it must be present) and revenuecat-status (the CLI ships a starter prompt for it). Genericize the help text that hardcoded "four skills". Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
bugbot run |
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 9b01de4. Configure here.
| var defaultToolkitSkills = []string{ | ||
| "create-revenuecat-project", | ||
| "integrate-revenuecat", | ||
| "revenuecat-cli", |
There was a problem hiding this comment.
Unknown default skill name
High Severity
revenuecat-cli was added to defaultToolkitSkills, but that skill does not appear in the published RevenueCat/ai-toolkit catalog (unlike sibling additions such as revenuecat-status). Default rc skills install and rc setup both pass this list to npx skills add --skill …, so an unknown skill name can fail the entire install and block onboarding.
Reviewed by Cursor Bugbot for commit 9b01de4. Configure here.
There was a problem hiding this comment.
False positive: revenuecat-cli does exist in the RevenueCat/ai-toolkit catalog on main (added in ai-toolkit#51, merged earlier today) — verified revenuecat/skills/revenuecat-cli/SKILL.md is present. rc skills install pulls from the toolkit default branch, so the name resolves. Bugbot was working from a snapshot taken before #51 merged.
Puts
rc setup(the account/app onboarding flow) on the same guided rail asrc setup google/rc setup apple, so all three setup experiences share the ┌ │ ◇ └ gutter, rail-native pickers, and a spinner ledger.Reuses the existing Flow / Ledger / rail-prompt primitives — no new UI machinery. The interface is the only thing that changed:
Edge cases kept intact:
projects usepicker) run in their own UI; the rail pauses and resumescodex mcp login) stay on the rail--yesstill skips the launch gate;--quietand--no-colorflow through--jsonoutput is unchanged (still the stage-aware prompt)Note
Medium Risk
Changes interactive onboarding and OAuth signup output paths; auth still uses the same provisioning flow but setup now owns signup UX and credential handling on the rail.
Overview
rc setupnow runs on the same guided rail (Flow: intro, steps, receipts, ledger) as the other setup flows, replacing scatteredrt.Outtitles and huh forms with rail-native Select, Confirm, Input, and Password prompts.Onboarding behavior changes: a roadmap preview at intro; account identity receipts; signup is collected on the rail via
setupSignupinstead of delegating torc auth signup(which would print off-rail).signupWithOAuthgains afromSetupflag and returns(passwordSaved, error)so setup can drive account creation under a ledger while suppressing duplicate success/hint chatter.--autonomyand--skills-scopeflags replace interactive autonomy/skills pickers (defaults shown as receipts;--skills-scopedefault global).Launch path: skills install uses a ledger step; MCP configuration returns an
mcpResultfor rail narration instead of writing directly to output; launch confirm respects--yesviaAssumeYes. Copyable agent prompt still prints raw to stdout (no gutter).Toolkit defaults:
defaultToolkitSkillsaddsrevenuecat-cliandrevenuecat-status.TUI:
Flow.Hint, optional scriptedinon rail prompts for tests; new tests guard gutter consistency and scripted prompts.Reviewed by Cursor Bugbot for commit 9b01de4. Bugbot is set up for automated code reviews on this repo. Configure here.