[jgoux]

Without the stack opt-in, supabase start --runtime native reaches the legacy start. That command fails with a bare Unrecognized flag: --runtime, which gives no clue that the flag belongs to the stack backend.

When the legacy backend rejects a flag that the stack version of the same command accepts, the error now names the opt-in:

Unrecognized flag: --runtime in command supabase start

  Hint: --runtime needs the stack backend. Set SUPABASE_EXPERIMENTAL_STACK=1, or set stack = true under [experimental] in your project config and leave SUPABASE_EXPERIMENTAL_STACK unset.

• The CLI builds the root that the stack opt-in would select and passes it to the parse-error formatters. The hint appears only when that root's command at the same path accepts the rejected flag.
• This covers every stack-only flag on start, stop and status without a hand-kept list. A flag that neither backend accepts gets no hint.
• Text and JSON output both carry the hint. Error codes are unchanged.

[github-actions]

CLI benchmark (Lightweight Commands): no changes found

Commit 1b781f5 against develop at df9a973 · 10 runs each · report only, never blocks merging

Command	Workload	Before	After	Change
supabase --version	version	157 ms	152 ms	no real change
supabase --help	help	192 ms	193 ms	no real change
supabase init	init	178 ms	174 ms	no real change
supabase migration new bench	migration-new	187 ms	189 ms	no real change
supabase config convert	config-convert	197 ms	196 ms	no real change
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	207 ms	214 ms	no real change
supabase projects list	projects-list	195 ms	194 ms	no real change
supabase functions list --project-ref replayfixtureprojref	functions-list	195 ms	195 ms	no real change

API calls: unchanged for all 8 commands.

Binary size: 135.0 MB, up 4.0 KB from 135.0 MB.

Details: timing ranges, API calls per command, slowest phases

Timing ranges

Command	Workload	Before range	After range	Change	Verdict
supabase --version	version	150-165 ms	150-161 ms	-5 ms (-3%)	within noise
supabase --help	help	168-229 ms	178-215 ms	+1 ms (+0%)	within noise
supabase init	init	167-210 ms	164-244 ms	-4 ms (-2%)	within noise
supabase migration new bench	migration-new	164-200 ms	161-216 ms	+1 ms (+1%)	within noise
supabase config convert	config-convert	185-207 ms	184-212 ms	-1 ms (0%)	within noise
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	194-219 ms	204-226 ms	+7 ms (+4%)	within noise
supabase projects list	projects-list	189-203 ms	182-203 ms	-1 ms (0%)	within noise
supabase functions list --project-ref replayfixtureprojref	functions-list	181-209 ms	173-215 ms	+0 ms (+0%)	within noise

API calls per command

Command	Workload	Before	After	Endpoint changes
supabase --version	version	0	0
supabase --help	help	0	0
supabase init	init	0	0
supabase migration new bench	migration-new	0	0
supabase config convert	config-convert	0	0
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	2	2
supabase projects list	projects-list	1	1
supabase functions list --project-ref replayfixtureprojref	functions-list	2	2

Slowest phases

No command moved beyond noise, so there are no phase deltas.

Workflow run

Comment /bench to run the lightweight benchmark (this comment) again on the latest commit, or /bench full to also run the integration-heavy benchmark, which posts in its own comment.

[github-actions]

🤖 AI Review

Verified all three findings and merged the overlapping test findings into one. Two minor concerns are confirmed: misleading config advice when an environment override disables stack, and an acceptance test that can pass despite unrelated failures.

Findings

Severity	Location	Category	Sources	Claim
🟡 MINOR	apps/cli/src/shared/cli/subcommand-flag-suggestions.ts:334	error-messages	claude	When SUPABASE_EXPERIMENTAL_STACK=0 is set, the hint incorrectly offers setting stack = true in config.toml as an effective alternative. The environment override prevents that config edit from enabling stack.
🟡 MINOR	apps/cli/src/cli/stack-flag-opt-in-hint.integration.test.ts:112	test-coverage	claude+codex	The stack-backend acceptance test checks only that the message lacks 'Unrecognized flag', so unrelated failures satisfy the assertion without establishing successful acceptance.

Stats

Claude findings: 2 · Codex findings: 1 · Confirmed: 2 · Refuted: 0 · Uncertain: 0

---

Models: claude-opus-5-5 + gpt-6.1-sol · Trigger: auto · Workflow run

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