feat(cli): add a Lovable handoff target to plan and impl - #871
Merged
Conversation
Lovable's agent runs in Lovable's cloud and only sees the repo through GitHub sync, so without repo-local guidance it answers CipherStash questions from stale training data (the pre-EQL-v3 "needs a Postgres extension and superuser" story) and talks users out of a supported Supabase setup. The new --target lovable (and picker entry) writes the same AGENTS.md as the editor-agent handoff — doctrine plus the per-integration skills inlined — with Lovable-specific next steps: commit and push the generated files, then add a Knowledge note in the Lovable project settings pointing the agent at AGENTS.md and .cipherstash/setup-prompt.md, since Lovable does not auto-load AGENTS.md the way Cursor or Windsurf do.
🦋 Changeset detectedLatest commit: 52104d7 The changes in this PR will be included in the next version bump. This PR includes changesets to release 11 packages
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
📝 WalkthroughWalkthroughAdds ChangesLovable handoff
Estimated code review effort: 3 (Moderate) | ~20 minutes Sequence Diagram(s)sequenceDiagram
participant User
participant StashCLI
participant ProjectFiles
participant Lovable
User->>StashCLI: Run plan or impl with --target lovable
StashCLI->>ProjectFiles: Write AGENTS.md and setup-prompt artifacts
StashCLI->>User: Show GitHub sync and Knowledge instructions
User->>Lovable: Configure project Knowledge
Lovable->>ProjectFiles: Reference AGENTS.md and setup-prompt guidance
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/cli/src/commands/impl/steps/handoff-lovable.ts`:
- Around line 42-64: Preserve the writeAgentsMd() result when generating the
Lovable handoff and setup prompt. In
packages/cli/src/commands/impl/steps/handoff-lovable.ts lines 42-64, pass an
explicit AGENTS.md write-status value and show recovery guidance instead of
instructing users to commit AGENTS.md when writing fails. In
packages/cli/src/commands/init/lib/setup-prompt.ts lines 254-276, update
skillsLoadedLines() to claim durable rules are in AGENTS.md only when artifact
generation succeeded. Add a regression test covering failed AGENTS.md writing.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: f0fd1209-0848-4fe7-a9b3-d7586324238f
📒 Files selected for processing (10)
.changeset/lovable-handoff-target.mdpackages/cli/src/cli/registry.tspackages/cli/src/commands/impl/__tests__/how-to-proceed.test.tspackages/cli/src/commands/impl/steps/handoff-lovable.tspackages/cli/src/commands/impl/steps/how-to-proceed.tspackages/cli/src/commands/init/lib/setup-prompt.tspackages/cli/src/commands/init/types.tspackages/cli/src/commands/plan/index.tspackages/cli/tests/e2e/impl-non-tty.e2e.test.tsskills/stash-cli/SKILL.md
Review follow-up on the Lovable handoff. When writeAgentsMd() fails, the handoff note now shows recovery guidance instead of telling the user to commit a file that was never written — in both the Lovable and editor-agent (agents-md) steps, which shared the flaw. skillsLoadedLines() keyed its 'durable rules are in AGENTS.md' claim on any AGENTS.md-writing handoff, but for agents-md and lovable the skills are inlined into AGENTS.md itself — all skills failing means the file was not written. Only codex writes its doctrine separately, so only codex keeps that pointer on the all-failed path.
freshtonic
approved these changes
Aug 10, 2026
freshtonic
left a comment
freshtonic
left a comment
Contributor
There was a problem hiding this comment.
Reviewed the full diff plus the surrounding code on the branch (target validation, the setup-prompt renderer, write-context.ts, and the plan/impl outro paths). Approving — this is a clean, well-scoped addition.
What holds up well:
--target lovableneeds no bespoke validation:planandimplboth validate againstHANDOFF_CHOICES, and the non-TTY hints ininit/plan/implall derive from it too — theplanoutro's switch from a hard-coded string toHANDOFF_CHOICES.join('|')closes the last one. The e2e assertion pins the rendered list.- The deferred-handoff behaviour falls out correctly for free:
handoffLovableStepnever setsagentLaunched, soplantakes the "No plan drafted yet, exit 0" path exactly as documented in the updatedskills/stash-cli/SKILL.md. - The honesty fixes are the strongest part.
skillsLoadedLinesnarrowing the all-failed AGENTS.md pointer tohandoff === 'codex'is a genuine bug fix for the pre-existingagents-mdpath (all-failed there meanswriteAgentsMditself failed, so the old prompt pointed at a nonexistent file), and both handoff notes now branch onwritten. The new tests cover exactly these branches, including the codex contrast case. - Repo hygiene is all in order:
stashminor changeset, SKILL.md updated against the manifest, registry help text, and the test updates match.
One doc nit (fine as a fast-follow):
packages/cli/README.md:27still says--target <claude-code|codex|agents-md|wizard>— the same stale list this PR fixes in the registry and SKILL.md. (The CHANGELOG hit is historical and correctly left alone.)
Observed but not worth acting on:
- The stripped-build + unwritable-AGENTS.md combination (
installed/inlined/failedall empty withwritten === false) would still claim "the durable rules are inAGENTS.md" for agents-md/lovable — but that requires two independent failures at once and predates this PR. - The success/failure note duplication between
handoff-agents-md.tsandhandoff-lovable.tsis acceptable; the cross-referencing comments make the pairing discoverable.
cipherstash-bot
left a comment
cipherstash-bot
left a comment
There was a problem hiding this comment.
Verdict
Solid PR — most of the Lovable handoff target is well covered (setup-prompt cases per mode, five-target ordering in how-to-proceed.test.ts, non-TTY e2e --target string, agents-md honesty branch source). The gaps are all test coverage, not correctness: the new execution path (handoffLovableStep.run) and one changed help surface (plan --help) are unverified. No crypto/security concerns — pure CLI handoff wiring.
Review stats
| Source | Model | Raw | Survived |
|---|---|---|---|
| claude | claude-opus-4-8 [test-gap] | 2 | 2 |
| codex | gpt-5.5 [test-gap] | 3 | 3 |
After de-dup: 4 distinct findings kept (the lovable-step gap was raised by both). Cross-model overlap: 1 kept finding corroborated by 2+ models.
…TS.md honesty branch Coverage-bot follow-up on the Lovable handoff. Four gaps, all in new or changed code that CI would have shipped green: - handoff-lovable.test.ts — neither arm of the writeAgentsMd success/ failure split was exercised, so a regression could record skills as inlined after the write failed, or tell the user to commit a file that was never written. - handoff-agents-md.test.ts — the step had no test at all, and this branch gained the same honesty ternary. - how-to-proceed-dispatch.test.ts — buildOptions ordering was covered but not the dispatch arms. Table-driven off HANDOFF_CHOICES so a target that reaches the picker without a dispatch arm fails here. - The --help e2e now loops over impl and plan; each carries its own --target descriptor, so asserting only impl let a dropped target ship silently from plan. Mutation-checked: rerouting the lovable dispatch arm and forcing the note ternary true each fail the new tests.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts (1)
28-28: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winReplace the repeated type-erasing
InitStatefixtures.Both tests force partial objects through
as unknown as InitState. This bypasses theInitStatecontract and duplicates the exception. Use a typed shared test fixture, or document one intentional coercion at a helper boundary.
packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts#L28-L28: replace the local type-erasing fixture.packages/cli/src/commands/impl/steps/__tests__/handoff-agents-md.test.ts#L27-L27: reuse the typed fixture.As per coding guidelines, source type-erasing assertions must use narrowing, a specific assertion, or a deliberate suppression with a reason. Based on learnings, minimize type-erasing assertions in
__tests__and centralize them when practical.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts` at line 28, Replace the repeated type-erasing InitState fixtures with one shared, properly typed test fixture or helper, centralizing any necessary coercion with an explicit reason. Update packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts#L28-L28 to use it, and update packages/cli/src/commands/impl/steps/__tests__/handoff-agents-md.test.ts#L27-L27 to reuse the same fixture; preserve each test’s existing integration value.Sources: Coding guidelines, Learnings
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts`:
- Line 28: Replace the repeated type-erasing InitState fixtures with one shared,
properly typed test fixture or helper, centralizing any necessary coercion with
an explicit reason. Update
packages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.ts#L28-L28
to use it, and update
packages/cli/src/commands/impl/steps/__tests__/handoff-agents-md.test.ts#L27-L27
to reuse the same fixture; preserve each test’s existing integration value.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 914840e8-cfe3-4c01-a4c2-8e4c24af5d55
📒 Files selected for processing (4)
packages/cli/src/commands/impl/steps/__tests__/handoff-agents-md.test.tspackages/cli/src/commands/impl/steps/__tests__/handoff-lovable.test.tspackages/cli/src/commands/impl/steps/__tests__/how-to-proceed-dispatch.test.tspackages/cli/tests/e2e/impl-non-tty.e2e.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- packages/cli/tests/e2e/impl-non-tty.e2e.test.ts
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #873.
Why
A customer hit a wall setting up CipherStash in Lovable. Lovable's agent — with no repo-local guidance available — answered from stale training data: it claimed CipherStash needs a Postgres extension (EQL) plus ZeroKMS "installed on the managed database", concluded a real setup was impossible, and built a hand-rolled AES-GCM + deterministic-HMAC scheme instead. Every claim is wrong today: EQL v3 is plain SQL, installs as a non-superuser on Supabase (Lovable's backend), and ZeroKMS is hosted — nothing installs in the database.
Nothing in
stash inittargeted Lovable, so there was no way to get correct guidance in front of its agent.What
Adds a
lovablehandoff target tostash plan/stash impl(--target lovableplus a new agent-target picker entry).AGENTS.mdwith the doctrine plus the per-integration skills inlined, and the usual.cipherstash/context.json+.cipherstash/setup-prompt.md.AGENTS.mdthe way Cursor/Windsurf do — so the guidance is to commit and push, then add a Knowledge note in the Lovable project settings pointing the agent atAGENTS.mdand.cipherstash/setup-prompt.md.plan's non-TTY outro now derives its target list fromHANDOFF_CHOICESinstead of a hard-coded string.skills/stash-cli/SKILL.md, and tests updated;stashminor changeset included.Verification
pnpm --filter stash test— 1207 passedpnpm --filter stash test:e2e— 100 passedpnpm run code:check— no errorsstash manifest --jsonresolves the new--targetvalue (skills doc checked against it)Summary by CodeRabbit
New Features
Bug Fixes
Documentation
Summary by CodeRabbit
planandimplcommands.