Evaluator path · public install

Install it. Inspect it. Rebuild it.

This path tests the public install command before trusting the README, then points to the immutable artifacts and source gates behind each public claim. No private credential, hosted service, or optional learning extension is required.

Version: 4.1.0
Core skills: 23
Engine revision: f563a15b158e…
Maintained files: 80

01 · Clean install

Start in a disposable repository.

Use Node.js 22, Git, and network access. Starting from an empty Git repository prevents an existing global skill or cached project setup from masking a packaging bug.

macOS · Linux · PowerShell: use npx.cmd~3 min

$ mkdir styleseed-evaluation && cd styleseed-evaluation
$ git init
$ npx skills add bitjaru/styleseed --agent codex --yes --copy
$ npx skills list --json --agent codex

02 · Follow the receipts

Every claim should end at inspectable evidence.

01 · Release

Inspect v4.1.0 as an immutable artifact

The release publishes the core archive, inventory, manifest, checksums, source revision, and the exact benchmark boundary used for this version.

Open release assets

02 · Automation

Read the checks that protect main

Engine contracts, runtime behavior, public claims, generated-file drift, markdown paths, plugin boundaries, palettes, and the production site build run in CI.

Inspect GitHub Actions

03 · Windows

Reproduce the PowerShell path

The Windows guide records the tested environment, npx.cmd fallback, expected 23-skill layout, and the distinction between the core installer and optional repository extensions.

Open Windows evidence

04 · Benchmark

Separate historical evidence from new claims

BENCH-V1 exposes its 120-cell result and raw evidence. Version 4.1 adds no new performance or superiority claim; the release records that waiver explicitly.

Read BENCH-V1

03 · Source verification

Rebuild the repository, not the story.

The full source path regenerates catalogs, runs the runtime and public-claim tests, validates the engine, checks links, and builds the production Next.js site. The exact command list lives in the evaluator guide so it can evolve with CI.

Run the full evaluator guide Review SBOM and licenses

Evidence boundary: a green local run is not a GitHub Actions run, an Actions run is not a production deployment, and BENCH-V1 is historical benchmark evidence rather than a new v4.1 performance claim.