refactor(cli): cover bootstrap with effect lint (CLI-2439)

[7ttp]

TL;DR

brings the supabase bootstrap command family under the effect lint

whats introduced?

effect lint applied to bootstrap:

• one allow list entry in .oxlintrc.effect.json
• the ambient SUPABASE_DB_PASSWORD and GITHUB_TOKEN reads move onto CommandSettings as redacted optional fields, and the raw SUPABASE_WORKDIR value rides along verbatim, all captured the same way as the access token
• the .env.example parse failure becomes a typed error with its telemetry tag registered
• the samples listing decodes through a schema codec
• integration tests pin the settings sourced create password and the env configured workdir

ref:

• closes: CLI-2439

[github-actions]

🤖 AI Review

Codex reported no findings. Of Claude's eight findings, six are confirmed and two refuted. Confirmed issues comprise one untyped template-decoding defect, two test-coverage gaps, one test-isolation concern, one misleading comment, and one non-total Effect thunk. No critical or major defects were verified.

Findings

Severity	Location	Category	Sources	Claim
🟡 MINOR	apps/cli/src/commands/bootstrap/bootstrap.templates.ts:138	error-handling	claude	The samples codec accepts any JSON value, after which property access on null can throw an untyped defect instead of BootstrapTemplateListError.
🟡 MINOR	apps/cli/src/commands/bootstrap/bootstrap.templates.ts:78	test-coverage	claude	The settings-sourced authorization header and schema-based samples decoding have no direct service-level tests.
🟡 MINOR	apps/cli/src/commands/bootstrap/bootstrap.workdir-cache.integration.test.ts:232	test-isolation	claude	The test constructs the real command settings layer without fully isolating process environment variables, allowing ambient profile or credential values to leak into it.
🟡 MINOR	apps/cli/src/commands/bootstrap/bootstrap.handler.ts:155	effect-idiom	claude	process.chdir is a throwing operation wrapped in Effect.sync, so a filesystem race or permission failure becomes a defect rather than a typed bootstrap failure.
⚪ NIT	apps/cli/src/commands/bootstrap/bootstrap.integration.test.ts:346	documentation	claude	The comment incorrectly attributes prompting to explicitWorkdir, although the handler does not consult that field.
⚪ NIT	apps/cli/src/commands/bootstrap/bootstrap.integration.test.ts:355	test-coverage	claude	The test named as using SUPABASE_WORKDIR without prompting does not assert that no prompt occurred.

Findings outside the diff

• 🟡 MINOR apps/cli/src/commands/bootstrap/bootstrap.handler.ts:155 — process.chdir is a throwing operation wrapped in Effect.sync, so a filesystem race or permission failure becomes a defect rather than a typed bootstrap failure.

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

• apps/cli/src/config/command-settings.layer.ts:147 (consistency): Capturing DB and GitHub credentials in CommandSettings conflicts with remaining lazy environment reads and creates contradictory sources of truth.
  Refuted: The code defines distinct lifecycles rather than contradictory sources: bootstrap consumes captured settings, while db-config intentionally performs per-invocation resolution. During bootstrap, resolveLinkedConn receives Option.some(created.dbPassword), so its ambient fallback cannot override the captured/created password.
• apps/cli/src/commands/bootstrap/bootstrap.suggest.ts:1 (style): Using import { type Path } creates an inappropriate value import and violates an established repository convention requiring import type.
  Refuted: The inline type modifier makes Path type-only and produces no runtime value import. Other files' use of import type does not establish a documented rule against this semantically equivalent syntax.

Stats

Claude findings: 8 · Codex findings: 0 · Confirmed: 6 · Refuted: 2 · Uncertain: 0

---

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

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