feat(cli): add experimental stack stop

[jgoux]

supabase experimental stack stop stops the current or named managed stack, or an explicit stack ID, while preserving its identity and data. Selection works without loading project configuration, including when the config file has been removed or changed.

Use the stack package’s lifecycle API and report whether a stack was found in both human and structured output. This PR builds on the experimental stack start integration.

[github-actions]

Supabase CLI preview

npx --yes https://pkg.pr.new/supabase/cli/supabase@865c92fd59ff639193d4a3aa299c1f2af3223556

Preview package for commit 865c92f.

[jgoux]

/ai-review

[jgoux]

/ai-review

[avallete]

Approve with comments. No blockers: the command meets the stop contract (config-free targeting, preserve data, never destroy). Non-blocking notes on cleanup-error actionability, test names vs asserts, and sharing the small targeting validations with start.

[github-actions]

🤖 AI Review

Only Claude's independent review was available; the Codex pass did not complete. Eight of Claude's ten findings are confirmed, covering shared targeting duplication, conditional documentation inaccuracy, a compiled-boundary test gap, and several maintainability/test-quality nits. Two findings are refuted: StackRuntimeMismatchError is not emitted by the production openStack path, and the documented catch-all exit-1 condition already covers both flagged validation errors. No critical or major defect was verified.

Findings

Severity	Location	Category	Sources	Claim
🟡 MINOR	apps/cli/src/commands/experimental/stack/stop/stop.handler.ts:96	code-duplication	claude	The stop handler duplicates stack-target validation already used by the stack command family instead of sharing it from the family root.
🟡 MINOR	apps/cli/src/commands/experimental/stack/stop/SIDE_EFFECTS.md:11	documentation	claude	The blanket statement that a missing project config does not prevent stopping is inaccurate for implicit or named targeting when workdir resolution changes the project root.
🟡 MINOR	apps/cli/src/commands/experimental/stack/stop/stop.integration.test.ts:82	test-coverage	claude	No compiled CLI test exercises experimental stack stop through the real stack API and supervisor teardown path.
⚪ NIT	apps/cli/src/commands/experimental/stack/stop/stop.handler.ts:33	dead-code	claude	The non-stack-error branch in stopError is unreachable through its statically typed call sites.
⚪ NIT	apps/cli/src/commands/experimental/stack/stop/stop.handler.ts:15	maintainability	claude	ExperimentalStackStopFlags is manually duplicated instead of being inferred from the command configuration.
⚪ NIT	apps/cli/src/commands/experimental/stack/stop/stop.integration.test.ts:167	test-quality	claude	Several stop integration-test names claim semantics their fixtures do not exercise.
⚪ NIT	apps/cli/src/shared/cli/hidden-flag.unit.test.ts:182	test-quality	claude	The two useDocker assertions cannot distinguish parsed flag values from command defaults.
⚪ NIT	apps/cli/src/shared/cli/hidden-flag.unit.test.ts:126	scope	claude	The hidden-flag test rewrite is unrelated to stack stop and removes a real-handler GoProxy forwarding assertion, although that contract remains covered elsewhere.

Refuted findings (kept for transparency, not posted as review comments)

• apps/cli/src/commands/experimental/stack/stop/stop.handler.ts:58 (error-classification): StackRuntimeMismatchError is a reachable openStack failure but is classified as unknown instead of actionable flags.
  Refuted: Although OpenStackError includes StackRuntimeMismatchError, the production openStack implementation cannot emit it. Its state read has a narrower failure set and makeHandle is non-failing. A mismatch returned during stop RPC is also narrowed by the package to StackLifecycleConflictError at EffectStack.ts:282-287.
• apps/cli/src/commands/experimental/stack/stop/SIDE_EFFECTS.md:24 (documentation): The exit-status documentation omits the exit-1 paths for legacy output-flag rejection and mutually exclusive target flags.
  Refuted: Both cited paths are typed ExperimentalStackStopError failures and therefore fall under the document's explicit catch-all "any typed stop failure" exit-1 condition. Their behavior could be documented more specifically, but the exit-status enumeration does not omit them.

Stats

Claude findings: 10 · Codex findings: 0 · Confirmed: 8 · Refuted: 2 · Uncertain: 0

---

Models: claude-opus-5 + gpt-5.6-sol · Trigger: manual · Workflow run

This review runs once per PR. A maintainer can request another with a /ai-review comment.