feat(compute): accept --cpu 16 (new 16vcpu-16gb rung)

[jwfing]

What

The platform adds a 16vcpu-16gb shared-cpu rung (SHARED_CPU_LADDER = 1, 2, 4, 6, 8, 16) and raises the pro/team/enterprise compute ceiling to 16 vCPU / 16384 MB. Free stays at 4/4. This PR updates the CLI's local grid to match:

• src/commands/compute.ts CPU_SIZES: adds 16, so parseCpu accepts --cpu 16
• src/index.ts: both --cpu help strings now read (provider sizes: 1, 2, 4, 6, 8, 16)
• test/limits.test.ts: the accepted grid now includes 16. Added 12 and 32 as off-grid cases that must still be rejected.

No other hardcoded ladder or 8 vCPU / 8 GB cap text found in src/, test/, README.md. skills/ main has no copy of the size list either.

Pairs with the platform PR: InsForge/instacloud-platform#609. Merge or release this only after that ships. Until then the server rejects --cpu 16 with its own error.

Tests

• npm run typecheck: clean
• npm test: 101 files / 2008 tests passed, 0 failed
• npm run build: clean

🤖 Generated with Claude Code

---

Summary by cubic

Adds 16 to the local vCPU grid so the CLI accepts --cpu 16, matching the new 16vcpu-16gb shared-cpu rung on the platform. The help strings for --cpu now advertise the new size, and the tests cover 12 and 32 as off-grid values that must still be rejected.

Do not release this before the platform PR ships—until then the server rejects --cpu 16.

^{Written for commit 3f6765a. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[jwfing]

Summary
The supplied diff consistently adds 16 to CPU validation, both help strings, and parser tests. No blocking correctness, security, or performance issues were identified. Review confidence is limited because sandbox initialization failed, preventing repository inspection and test execution.

Requirements context
Assessed against the supplied PR description, diff, and AGENTS.md instructions. The stated intent is to accept --cpu 16 while retaining rejection of off-grid sizes, with merge/release gated on platform PR #609 shipping. No additional matching requirements documents could be verified: local file access failed and GitHub fetches were unsuccessful. The development guide and agent-facing reference documentation could not be inspected.

Findings

Critical
(none)

Suggestion
(none)

Information

• test/limits.test.ts:48: Software engineering: focused regression coverage - The changed tests follow the existing loop-based pattern, add acceptance of 16, and add rejection of 12 and 32 while preserving previous cases. The implementation retains the existing validation and error-handling style.
• Functionality: Validation and advertised sizes agree - The supplied diff adds 16 to CPU_SIZES and both compute and managed-database help strings. Existing accepted sizes remain available. The stated platform-first rollout requirement still applies; platform deployment status could not be verified.
• Security: No security regression visible in the diff - Validation remains a numeric allowlist. The change introduces no dependencies, sensitive logging, SQL or shell execution, or authorization changes. Server-side plan enforcement was outside the accessible review evidence.
• Performance: No meaningful performance impact - The only runtime change expands a fixed allowlist from five to six entries. No additional I/O, unbounded work, or significant allocation is introduced.

Tests
No tests ran. The initial read-only shell command failed before execution with a bwrap namespace-permission error, preventing working-tree cross-checks and surrounding-code inspection. The PR reports successful typecheck, 2008 tests across 101 files, and build; these results were not independently verified.

Verdict
approved: No Critical findings were identified in the supplied diff, so the verdict is approved under the required rule, subject to the documented review limitations.

[Fermionic-Lyu]

LGTM, Approved.