[ntucker]

Requested by Nathaniel · project thread

Motivation

List the data-client plugin in the official directories (Claude's plugin directory, Cursor Marketplace, OpenAI's ChatGPT/Codex plugin directory) so people find it without knowing the repo. Each submission needs Nathaniel's account; this PR makes the repo pass their checks.

Solution

• Claude directory: it submits a folder holding .claude-plugin/plugin.json, with everything the plugin uses inside it. .agents/skills (already Cursor's plugin root) now gets that manifest, written by plugin-dir.mjs (renamed from cursor.mjs) from marketplace.json like Cursor's copy, and checked for drift by the same test. It has no version, so every commit to master is an update. claude plugin validate ./.agents/skills passes (only warning: no version, intended).
• Listing README at .agents/skills/README.md: the directory shows it as the description and blocks under 40 words (a test guards that); Cursor's checklist wants one too. It says what the plugin runs and sends (nothing of its own; installs only when asked), which the directory's security scan looks for, and has install lines for all three agents.
• Logo: the directory rejected assets/logo.svg as an icon because of its inline <style>; now plain fill.
• OpenAI: its directory uploads a ZIP with skills at skills/<name>/SKILL.md, not a repo. node .claude-plugin/openai-zip.mjs builds it (validated with claude plugin validate after unzipping).
• Cursor's own validate-template.mjs passes except for the git-subdir object source, which we keep on purpose (see .claude-plugin/README.md).

No package or docs-page changes; the Installation page tabs switch to marketplace commands once the listings are live.

Open questions

Claude's directory refuses repositories whose GitHub archive passes 50 MiB. This repo's is 47.3 MiB today, mostly website videos, so the next video would stop new plugin versions from being scanned. Discussed in the project thread.

---

Generated by Claude Code

---

Note

Low Risk
Documentation, manifest generation, tests, and CI around agent plugin packaging only—no runtime library or app behavior changes.

Overview
Prepares the data-client agent plugin for Claude, Cursor, and OpenAI marketplace submissions by making .agents/skills a self-contained plugin root with generated manifests, listing copy, and CI validation.

cursor.mjs is replaced by plugin-dir.mjs, which still syncs Cursor manifests from marketplace.json and now also writes .agents/skills/.claude-plugin/plugin.json (no version, so directory installs track every commit). Tests were renamed/extended (plugin-dir.test.mjs, README word-count guard, Claude icon path).

Adds a marketplace README under .agents/skills (install lines for all three agents, security/disclosure text) and simplifies logo.svg so Claude’s directory accepts the icon (plain fill, no inline <style>).

openai-zip.mjs builds data-client-plugin.zip for OpenAI’s upload flow (skills under skills/, semver from build time). agent-rules CI runs pinned claude plugin validate ./.agents/skills; the zip artifact is gitignored.

^{Reviewed by Cursor Bugbot for commit 1ca359f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[changeset-bot]

⚠️ No Changeset found

Latest commit: 1ca359f

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[chatgpt-codex-connector]

Codex Review Summary

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

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-10-21T04:20:24.202778Z We couldn’t complete this review. Comment @codex review to try again.	1ca359f	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.

[ntucker]

Staff engineer (Cursor agent): LGTM, plus one follow-up idea for the open question.

Architecture: This is the right shape. There are now three manifests (marketplace.json, the Cursor plugin.json, and the new Claude plugin.json), but the two copies are generated from one source by plugin-dir.mjs and the existing drift test covers both. The rename from cursor.mjs makes the generator honest about what it owns, and hoisting keywords/pluginSkills keeps the two copies from diverging. openai-zip.mjs is a manual upload step, so it can go stale, but that comes from OpenAI's ZIP-only intake. A script plus a README note is the minimal answer, and I wouldn't automate it until uploads turn out to be frequent.

Follow-up (no change needed here): the 50 MiB archive limit. Most of the archive is three website videos: website/static/videos/blog/data-client-nextjs.mp4 (18.9 MB), docs/normalization.mp4 (7.1 MB), and docs/normalization.webm (4.2 MB). MP4 and WebM barely compress, so together they're about 30 MB of the 47.3 MiB. GitHub's source archives honor export-ignore in .gitattributes, so a single line like website/static/videos/** export-ignore would drop them from the archive. That doesn't touch the repo, clones, or the Vercel docs build, which deploys from git, not from the archive. I think this beats moving the videos to LFS or a CDN, at least until the site needs videos outside that folder. I'd do it as a small separate PR once this lands, and check the archive size before and after.

What I validated vs. read: I read the full diff. I listed blob sizes on master through the GitHub API to confirm the video sizes above. I did not run claude plugin validate, the node tests, or openai-zip.mjs, and I didn't measure the archive. It's also an inference that Claude's directory measures the same archive GitHub serves (which applies export-ignore); that needs checking against a real archive download before relying on it.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.37%. Comparing base (bb0dda6) to head (1ca359f).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4443   +/-   ##
=======================================
  Coverage   98.37%   98.37%           
=======================================
  Files         161      161           
  Lines        3073     3073           
  Branches      604      604           
=======================================
  Hits         3023     3023           
  Misses         10       10           
  Partials       40       40           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ntucker]

Re the export-ignore follow-up: that one won't work. Claude's pre-submission checklist says to "keep export-ignore and export-subst out of every .gitattributes file", and when either is present validation stops with Couldn't validate that repository. So hiding the videos from the archive would block the submission outright, not shrink it. The options for staying under 50 MiB are now in the project thread: a mirror repo for the plugin, hosting the videos outside git, or a CI size guard. The choice has to be made before the Claude submission, because a submission's repo can't be changed afterwards.

---

Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4375575110

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.