refactor(plugin): migrate rank pool helpers to TypeScript

[kmbroai]

Summary

Move ranking pool planning, validation, and result merging to the bundled TypeScript helper.

Changes

• Preserve plan schemas, directory checks, deterministic assignments, shard validation, and receipt hashes over the original input bytes.
• Remove redundant aggregate assignment and slot checks; validate the canonical round-robin assignment once per worker.
• Read plans directly and use strict UTF-8 decoding while retaining BOM and UTF-16/32 support.
• Register the migrated commands and remove their Python commands and tests.

Testing

• Combined stack tip ba12023: both full SDK runs passed 2,656 tests with 50 skips and zero failures (seeds 12345 and 4201856736).
• The six affected helper suites passed 213 tests with four Windows-only skips; all 23 MCP test processes passed.
• CI compilation, plugin build, types, formatting, Ruff, portable source checks, and nine source-checker tests passed.
• Rust formatting and cross-target Clippy for the Windows x64 native proof passed. Local execution was Linux; hosted CI supplies platform runtime coverage.

Risk and rollout

Stacked on ranking shards. make-rank-pool-plan, validate-rank-worker, and validate-rank-pool move to launch_codex_security_mcp[.cmd] --helper, preserving arguments, defaults, and the existing worker bound. Malformed UTF-8 is rejected even in overwritten duplicate values; ASCII-escaped surrogates and valid UTF-8/16/32 plans retain their handling. Duplicate assignments report the existing round-robin error.

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
🔒 Security Review	✅ Completed 2026-09-11T00:04:56.336337Z	c46466b	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[mldangelo-oai]

Reviewed c60440ca2554c5bb3cfdea69f19763d291499c09 against its #841 base using GPT-6 Astra at ultra effort, including the PR deslop and critical review passes.

This preserves the existing callable pool API. Its immediate product value is limited: the current in-tree scan guidance excludes separate ranking-phase pools, and I found no production orchestrator or receipt parser beyond helper dispatch. That makes additional generic compatibility machinery difficult to justify.

I found no introduced correctness or security regression and identified two nonblocking reductions totaling 58 lines:

• Remove aggregate assignment checks and an unreachable slot check already implied by exact per-slot validation: 39 lines.
• Reuse strict UTF-8 decoding for pool-plan JSON while retaining BOM and UTF-16/32 support: 19 lines. This deliberately rejects malformed UTF-8 that an overwritten duplicate value can currently hide; the inline comment describes that compatibility change.

Verification beyond the complete-stack comparison with main:

• 163 focused tests passed, with 4 Windows-only skips; 29 remaining Python generator tests passed.
• 2,540 direct mutation/encoding cases passed their explicit oracles on each of Node 20.0, 22.13, 24 and 26, comparing the original with independent variants.
• 29 real CLI scenarios covered 304 invocations on each of Node 20 and 22. Original-byte receipt hashes, artifact bytes and modes were verified; invalid plans caused zero shard reads.
• The three disposable pool-test variants passed 162 tests. SDK/MCP builds and types, Ruff and portable source checks passed.

Keep original plan/output buffers for receipt hashing and the exact per-slot and shard-content validators. Coordinate the previously posted existence, native-directory and discovery simplifications through this caller.

Local OS execution was Linux. Exact-head CI supplies separate Windows, PowerShell, macOS and native-runtime evidence. The proposed changes need that platform CI when applied, and the synthetic combined stack needs it after restacking.

[kmbroai]

Applied the remaining reductions: aggregate assignment checks are removed, plan reads no longer precheck existence, and UTF-8 uses the shared strict decoder. Per-slot assignment validation, shard checks, BOM/UTF-16/32 support, and original-byte receipt hashing remain. The malformed-UTF-8 compatibility change is documented and tested.

Updated head: c46466b46e494958d47ea0dc499045cbff9e4d36. Hosted CI passes on this head, including Linux/macOS/Windows tests, native proofs, installed-package checks, plugin source contracts, types, and formatting. Codex Security Review also passes. The combined stack at 5bd1d1909ad99d2749fe8d3ec49a9a144ae2bac0 includes this head and has also passed platform CI.