refactor(plugin): migrate rank shard helpers to TypeScript

[kmbroai]

Summary

Move ranking shard creation, validation, and merging to the bundled TypeScript helper while preserving command options, shard layout, and output rows.

Changes

• Reuse the shared worklist parser and raw-byte directory enumeration.
• Generate canonical shard names and check their membership directly; preserve contiguous assignments, row partitions, paths, and area validation.
• Keep real Unicode/raw-filename tests and update the Windows native proof for the canonical-name diagnostic.
• 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 the Unix directory primitive. make-rank-shards, validate-rank-shard, and merge-rank-outputs move to launch_codex_security_mcp[.cmd] --helper, preserving options and defaults. Malformed shard names use the canonical-name diagnostic; missing inputs report their read 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 2026-09-11T00:06:36.444056Z	bc172b4	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 69bfaa291f22e795d7a4318c3ba202e55644762d against its #840 base using GPT-6 Astra at ultra effort, including the PR deslop and critical review passes.

The migration preserves the covered partition, worker-validation and filesystem contracts. I found no demonstrated correctness or security regression and identified two nonblocking simplifications totaling 42 lines:

• Generate canonical shard names and check membership instead of parsing and sorting numeric suffixes: 13 lines.
• Remove the inactive opendirSync preload from the Unicode/raw-name regression test: 29 lines, retaining its real filesystem assertions.

Verification beyond the complete-stack comparison with main:

• 89 focused shard/deep-review tests passed, with 2 Windows-only skips; 15 surviving Python pool-consumer tests passed.
• The discovery replacement passed 327 filesystem cases on each of Node 22.13, 24 and 26, including 10,001 canonical shards.
• 35 Python-base/head/variant scenarios and five pool compositions covered 326 helper invocations. Artifact bytes, modes, links, assignments and 15 independently recomputed worker receipts matched.
• An unconditional opendirSync trap recorded zero calls in both test scenarios; the shorter fixture passed.
• SDK/MCP types, builds, Ruff and portable source checks passed.

Malformed-name diagnostics intentionally use the canonical-name error. One downstream test asserting the previous wording needed that assertion updated; its focused rerun passed. Keep the exact partition/path/area checks and wide/raw-path handling. The native-export and input-existence suggestions remain on #840 and #839 to avoid duplicate findings.

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

[kmbroai]

Applied canonical shard-name generation and membership checks, carried the earlier removal of the unused preload, and updated the downstream diagnostic assertions and Windows native proof. The raw-filename, partition, path, area, and output-preservation checks remain.

Updated head: bc172b4428d9388b9f90fbd61fcac4ec8b8c436e. 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. GitHub still labels Windows shard 2 as in progress, although all its test and cleanup steps completed successfully; the workflow and required gates report success. The combined stack at 5bd1d1909ad99d2749fe8d3ec49a9a144ae2bac0 includes this head and has also passed platform CI.