Should you install Baoyu Skills?
Try a selected publishing workflow if you can review its scripts and test a disposable draft. Our sandbox installation succeeded, but the test command failed, and live publishing remains unverified. That makes a limited evaluation the supported next step. Teams requiring a green test gate should resolve the recorded failure before adoption.
The README recommends installing only the skills you need because bulk installation adds unnecessary agent context. Choose by the output you need: article artwork, editable diagrams or a publishing route. A large catalog does not establish that each workflow meets your requirements.
What our run established
Our run on September 27, 2026 tested commit 1567581 in our sandbox. The recorded environment was lab-node:22 with 3 CPUs and 8 GB of memory. This review uses the existing stored run; no new execution was performed.
Our sandbox's stored parser summaries record 236 passed, 2 failed and 238 total tests. The harness records command success, elapsed time, exit code and timeout status separately from parsed counts. Installation returned zero. Tests returned 1 after the recorded 20 seconds and did not time out. The build was skipped because no script was detected, so it has no passing result.
The retained raw test footer agrees with those parser counts and reports zero skipped tests. Its tail contains two generic test failed errors without the failed cases' names or diagnostic causes. We cannot assign those failures to a provider or publishing route. The summaries do not establish exhaustive unique test coverage, and no complete-log reconciliation was performed.
The record does not demonstrate a completed image-generation request, live account publishing, deployment or output quality. Test names mentioning browsers and account handling do not establish that a real publishing session succeeded. The MrKeyoor testing methodology explains how command outcomes and parsed counts reach reviews. Its current description supplies provenance context, not proof that this older run used today's procedure.
Installation worked; audit coverage remains unknown
The attached harness records a 50-second installation, 326 dependencies and a 203 MB installed footprint. The retained installer message reports a different package count and its own elapsed time. We preserve the harness values as the page's measurements rather than substituting the installer's figures or combining the two observations.
The stored npm audit summary records 11 advisories: 8 high, 2 moderate, 1 low and zero critical. These are recorded dependency advisories, not demonstrated exploits or a security assessment. No scannedPackages value is supplied, so scan coverage is unknown. The summary also does not identify which selected skills use affected dependencies. Inspect the advisory paths before deciding whether a workflow meets your requirements.
Start a WeChat article trial with the documented skill set
The README at the measured commit names baoyu-cover-image, baoyu-article-illustrator and baoyu-post-to-wechat as the usual minimal WeChat article set. Codex can discover complete skill directories under a project's .agents/skills. The publishing skill includes Markdown-to-WeChat HTML conversion, so that flow does not require a separate baoyu-markdown-to-html installation.
Choose the account route before configuring it. WeChat offers API, browser and remote API methods with different credential and host requirements. X and Weibo explicitly leave final publication to the user after scripts fill the browser editor. Do not extend that manual-publication claim to every WeChat route. Review account selection and generated content during your own trial.
Match the collection to the job
Anthropic Skills includes document creation and editing skills for DOCX, PDF, PPTX and XLSX files. Choose it for those document tasks or to study Claude skill examples. Its README calls the document skills source-available rather than open source and describes the repository as demonstration and educational material. It can complement Baoyu Skills; the supplied evidence does not establish it as a substitute for Baoyu's social-publishing routes.
Superpowers is a software development methodology built on composable agent skills. Choose it for planning, implementation, testing and code review. It can complement a publishing collection when your work also includes software development. Its documented engineering process does not establish equivalent publishing capabilities.
What the retained ratings mean
The editorial scale runs from 1 to 5. For setup, 1 means installation blocked evaluation, 3 means installation succeeded with workflow configuration remaining, and 5 requires demonstrated workflow setup. For documentation, 1 means essential instructions are absent, 3 means a basic route is explained, and 5 means installation scopes, examples and credential requirements are covered. Documentation ratings assess instructions rather than verified behavior.
For community, 1 means no participation evidence, 3 means a public contribution channel, 4 adds a credited contribution and documented fix, and 5 requires broader sustained participation evidence. The supplied commit credits a contributor and the release documents a fix; these do not measure adoption or response times. For maturity, 1 means setup could not proceed, 3 means installation worked with unresolved test or advisory results, and 5 requires stronger operational evidence than this run supplies. The retained ratings are setup 3, documentation 5, community 4 and maturity 3.