Skip to content

feat(marketing): source agent skills from superset-sh/skills, drop monorepo copy - #5919

Merged
saddlepaddle merged 3 commits into
mainfrom
skills-repo-migration
Jul 24, 2026
Merged

saddlepaddle merged 3 commits into
mainfrom
skills-repo-migration

Conversation

@saddlepaddle

@saddlepaddle saddlepaddle commented Jul 24, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

  • The official agent skills now live in the public superset-sh/skills repo (the convention official vendors use on skills.sh: anthropics/skills, cloudflare/skills, openai/skills). The previous private superset-sh/skills (sales-outreach commands) was renamed to superset-sh/sales-skills; GitHub redirects both old names.
  • The repo contains two skills — the existing superset CLI skill (moved verbatim) and a new superset-mcp skill (MCP setup one-liners, auth flow, tool catalog) — and is registered on skills.sh (npx skills add superset-sh/skills verified: both skills install).
  • This PR removes the monorepo copy at skills/superset/ and repoints consumers:
    • /.well-known/agent-skills/index.json + /{skill}/SKILL.md now fetch from the skills repo with 1h ISR (single dynamic route replaces the hardcoded one; serves both skills; unknown names 404; source outage → 503 with ISR serving stale-good in practice).
    • The Claude Code plugin (plugins/superset) symlinked the deleted path — it now vendors a real copy, since plugin distribution needs the file on disk.
    • llms.txt lists the skills repo + install command.

Why / Context

skills.sh only counts self-published official skills (published from the vendor's own org). Ours lived inside the monorepo, so scans saw only "1 community skill with potential relation." Follow-up to the agent-readiness series (#5910/#5912/#5917).

Manual QA Checklist

  • npx skills add superset-sh/skills --all installs both skills into .agents/skills/ (universal + Claude Code symlinks)
  • /.well-known/agent-skills/index.json lists both skills with descriptions + sha256 digests, sourced live from GitHub
  • /{skill}/SKILL.md serves text/markdown for both; unknown skill name → 404
  • bun run lint exits 0 (caught and fixed the plugin's broken symlink)

Known Limitations

  • Two copies of the CLI skill now exist (skills repo = canonical, plugin vendors one). If the skill changes, update the skills repo first and re-copy into plugins/superset/skills/superset/.
  • The skills repo carries ELv2 (copied from the monorepo). Most vendors ship skills under MIT since the content is meant to be copied into user projects — flip it if you prefer.

Summary by cubic

Move official agent skills to the public superset-sh/skills repo and drop the monorepo copy. Marketing now fetches skills from GitHub and serves both superset and superset-mcp; docs now point agents to the repo and install command.

  • New Features

    • /.well-known/agent-skills/{skill}/SKILL.md dynamically fetches from superset-sh/skills with 1h ISR; supports superset and superset-mcp (unknown → 404, source outage → 503).
    • /.well-known/agent-skills/index.json builds from live content with sha256 digests and includes repository: https://github.com/superset-sh/skills.
  • Migration

    • Removed skills/superset/ from the monorepo; the Claude Code plugin now vendors the CLI skill file on disk.
    • Updated developer resources and the agent integration docs to link the repo and install command: npx skills add superset-sh/skills.

Written for commit b27e072. Summary will update on new commits.

Review in cubic

Summary by CodeRabbit

  • New Features
    • Added discoverable agent skill listings plus Markdown skill endpoints for supported skills.
    • Updated the skill index to aggregate available skills dynamically and return clear availability vs. missing-skill responses.
    • Included enhanced discovery metadata and caching for improved performance.
  • Documentation
    • Added an “Agent Skills” section with setup guidance and the install command for the Superset CLI and MCP server skills.
    • Published detailed end-to-end superset CLI workflow documentation, including commands, environment variables, sessions, and troubleshooting.

…norepo copy

The official skills now live in the public superset-sh/skills repo
(registered on skills.sh via npx skills add superset-sh/skills; the old
private skills repo moved to sales-skills). The /.well-known/agent-skills
routes fetch SKILL.md files from that repo with 1h ISR instead of reading
a checked-in copy, and now serve both skills (superset CLI + superset-mcp).
llms.txt links the repo.

Claude-Session: https://claude.ai/code/session_01GbjRvApZBmhsSkfHzihp6x
The plugin symlinked ../../skills/superset, which the skills-repo
migration removed. superset-sh/skills is canonical; the plugin carries a
copy because plugin distribution needs the file on disk.

Claude-Session: https://claude.ai/code/session_01GbjRvApZBmhsSkfHzihp6x
@coderabbitai

coderabbitai Bot commented Jul 24, 2026 •

Copy link
Copy Markdown
Contributor

Review Change Stack

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 5a485430-ffed-4147-9958-b111da0ceb5c

📥 Commits

Reviewing files that changed from the base of the PR and between 3096dbb and b27e072.

📒 Files selected for processing (1)
  • apps/docs/content/docs/agent-integration.mdx

📝 Walkthrough

Walkthrough

Adds shared retrieval for supported agent skills, JSON and Markdown discovery endpoints, Superset CLI skill documentation, and installation links in marketing and integration docs.

Changes

Agent skills discovery

Layer / File(s) Summary
Shared skill source
apps/marketing/src/app/.well-known/agent-skills/skills-source.ts
Defines supported skill names, runtime validation, and revalidated retrieval from the raw repository.
Discovery endpoints
apps/marketing/src/app/.well-known/agent-skills/...
Aggregates available skills into index.json with metadata and digests, and serves validated Markdown responses with error and cache handling.
CLI skill documentation and entry points
plugins/superset/skills/superset/SKILL.md, apps/marketing/src/lib/llms.ts, apps/docs/content/docs/agent-integration.mdx
Documents Superset CLI workflows and commands, and adds Agent skills installation instructions to developer resources and integration docs.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Client
  participant IndexRoute
  participant SkillSource
  participant GitHubRawRepository
  Client->>IndexRoute: GET index.json
  IndexRoute->>SkillSource: Fetch supported skills
  SkillSource->>GitHubRawRepository: Request SKILL.md files
  GitHubRawRepository-->>SkillSource: Markdown content
  SkillSource-->>IndexRoute: Skill content or null
  IndexRoute-->>Client: JSON discovery index
Loading

Suggested reviewers: kitenite

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Title check ✅ Passed Conventional commit title clearly matches the migration to external agent skills and removal of the monorepo copy.
Description check ✅ Passed It covers what changed, why, testing, and a checklist, though the headings differ from the template and some checklist items are missing.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch skills-repo-migration

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@github-actions

github-actions Bot commented Jul 24, 2026 •

Copy link
Copy Markdown
Contributor

🚀 Preview Deployment

🔗 Preview Links

Service Status Link
Neon Database (Neon) ✅ View Branch
Vercel API (Vercel) ✅ Open Preview
Vercel Web (Vercel) ✅ Open Preview
Vercel Marketing (Vercel) ✅ Open Preview
Vercel Admin (Vercel) ✅ Open Preview
Vercel Docs (Vercel) ✅ Open Preview

Preview updates automatically with new commits

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 4

🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@apps/marketing/src/app/.well-known/agent-skills/skills-source.ts`:
- Around line 10-17: Update fetchSkill to catch both fetch and response.text()
failures, returning null for any remote error so callers retain their fallback
behavior. Add a bounded timeout to the remote request using the existing fetch
cancellation pattern before relying on next.revalidate, while preserving the
current null result for non-OK responses.

In `@plugins/superset/skills/superset/SKILL.md`:
- Line 18: Update the “Plan work” workflow instruction to use the complete
`superset tasks update <id-or-slug>` command, matching the existing `superset
tasks create` command.
- Line 122: Update the automation documentation around the project ID
prerequisite and the workspace examples to distinguish project-scoped and
workspace-scoped operations: clarify when --project is required versus optional,
and reconcile the statements about whether runs always create a workspace or
only create one when --workspace is omitted. Ensure the documented behavior is
consistent across the prerequisite text and examples.
- Line 11: Replace the direct curl-to-sh installation command in the Superset
CLI instructions with a safer installation method, preferably a package manager
or signed release. If no alternative exists, document downloading a pinned
script or release, verifying its integrity against a trusted checksum or
signature, and only then executing it.
🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

  • Push a commit to this branch (recommended)
  • Create a new PR with the fixes

ℹ️ Review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 33f5c92d-8fbf-4945-aeb5-ed6a44a9101c

📥 Commits

Reviewing files that changed from the base of the PR and between 549f950 and 3096dbb.

📒 Files selected for processing (7)
  • apps/marketing/src/app/.well-known/agent-skills/[skill]/SKILL.md/route.ts
  • apps/marketing/src/app/.well-known/agent-skills/index.json/route.ts
  • apps/marketing/src/app/.well-known/agent-skills/skills-source.ts
  • apps/marketing/src/app/.well-known/agent-skills/superset/SKILL.md/route.ts
  • apps/marketing/src/lib/llms.ts
  • plugins/superset/skills/superset
  • plugins/superset/skills/superset/SKILL.md
💤 Files with no reviewable changes (2)
  • plugins/superset/skills/superset
  • apps/marketing/src/app/.well-known/agent-skills/superset/SKILL.md/route.ts

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Caution

Inline review comments failed to post. This is likely due to GitHub's internal server error or limits when posting large numbers of comments. If you are seeing this consistently it is likely a permissions issue. Please check "Moderation" -> "Code review limits" under your organization settings.

Actionable comments posted: 4

🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@apps/marketing/src/app/.well-known/agent-skills/skills-source.ts`:
- Around line 10-17: Update fetchSkill to catch both fetch and response.text()
failures, returning null for any remote error so callers retain their fallback
behavior. Add a bounded timeout to the remote request using the existing fetch
cancellation pattern before relying on next.revalidate, while preserving the
current null result for non-OK responses.

In `@plugins/superset/skills/superset/SKILL.md`:
- Line 18: Update the “Plan work” workflow instruction to use the complete
`superset tasks update <id-or-slug>` command, matching the existing `superset
tasks create` command.
- Line 122: Update the automation documentation around the project ID
prerequisite and the workspace examples to distinguish project-scoped and
workspace-scoped operations: clarify when --project is required versus optional,
and reconcile the statements about whether runs always create a workspace or
only create one when --workspace is omitted. Ensure the documented behavior is
consistent across the prerequisite text and examples.
- Line 11: Replace the direct curl-to-sh installation command in the Superset
CLI instructions with a safer installation method, preferably a package manager
or signed release. If no alternative exists, document downloading a pinned
script or release, verifying its integrity against a trusted checksum or
signature, and only then executing it.
🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

  • Push a commit to this branch (recommended)
  • Create a new PR with the fixes

ℹ️ Review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 33f5c92d-8fbf-4945-aeb5-ed6a44a9101c

📥 Commits

Reviewing files that changed from the base of the PR and between 549f950 and 3096dbb.

📒 Files selected for processing (7)
  • apps/marketing/src/app/.well-known/agent-skills/[skill]/SKILL.md/route.ts
  • apps/marketing/src/app/.well-known/agent-skills/index.json/route.ts
  • apps/marketing/src/app/.well-known/agent-skills/skills-source.ts
  • apps/marketing/src/app/.well-known/agent-skills/superset/SKILL.md/route.ts
  • apps/marketing/src/lib/llms.ts
  • plugins/superset/skills/superset
  • plugins/superset/skills/superset/SKILL.md
💤 Files with no reviewable changes (2)
  • plugins/superset/skills/superset
  • apps/marketing/src/app/.well-known/agent-skills/superset/SKILL.md/route.ts
🛑 Comments failed to post (4)
apps/marketing/src/app/.well-known/agent-skills/skills-source.ts (1)

10-17: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "== locate file =="
git ls-files | rg 'apps/marketing/src/app/.well-known/agent-skills/skills-source\.ts|agent-skills' || true

echo "== file excerpt =="
if [ -f apps/marketing/src/app/.well-known/agent-skills/skills-source.ts ]; then
  cat -n apps/marketing/src/app/.well-known/agent-skills/skills-source.ts
fi

echo "== search fetchSkill usages =="
rg -n "fetchSkill|Promise\.all|agent-skills|AgentSkills" apps/marketing/src/app/.well-known apps/marketing/src 2>/dev/null || true

Repository: superset-sh/superset

Length of output: 2100


🏁 Script executed:

#!/bin/bash
set -euo pipefail

node - <<'JS'
(async () => {
  const results = await Promise.allSettled([
    (async () => {
      const response = await fetch("https://httpbin.org/status/404");
      if (!response.ok) return true;
      return await response.text();
    })().then(() => false),
    fetch("https://127.0.0.1:1").then(async response => response.ok ? await response.text() : null).then(() => false),
  ]);
  console.log("rejected responses:", results.some(r => r.status === "rejected"));
  console.log("reasons:", results.map(r => r.status === "rejected" ? r.reason && r.reason.name : null));
})().catch(e => {
  console.error("probe thrown:", e && e.name);
  process.exitCode = 1;
});
JS

Repository: superset-sh/superset

Length of output: 214


Handle rejected fetchSkill fetches before route composition.

fetchSkill only returns null for non-OK responses; network failures and response.text() read failures still reject. Since the index route awaits Promise.all([fetchSkill(...), ...]) and the per-skill route awaits fetchSkill(...) directly, those routes can surface a 500 instead of the fallback. Catch the fetch plus body read inside fetchSkill, return null on failure, and apply a bounded remote-timeout before relying on next.revalidate.

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@apps/marketing/src/app/.well-known/agent-skills/skills-source.ts` around
lines 10 - 17, Update fetchSkill to catch both fetch and response.text()
failures, returning null for any remote error so callers retain their fallback
behavior. Add a bounded timeout to the remote request using the existing fetch
cancellation pattern before relying on next.revalidate, while preserving the
current null result for non-OK responses.
plugins/superset/skills/superset/SKILL.md (3)

11-11: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

printf 'File context:\n'
sed -n '1,80p' plugins/superset/skills/superset/SKILL.md || true

printf '\nSearch for similar install commands nearby:\n'
rg -n 'curl .*superset|superset\.sh/cli/install|pip install|\.install\.sh' plugins/superset || true

printf '\nRepository files for superset skill/docs:\n'
fd -p 'plugins/superset/skills/superset|plugins/superset' . --type f | sed -n '1,120p'

Repository: superset-sh/superset

Length of output: 5156


Do not pipe an unpinned remote script directly into sh.

The install step downloads and executes mutable upstream code without inspection or integrity verification. Prefer a package manager or signed release; otherwise document a download-then-verify-pinned-checksum flow before execution.

🧰 Tools
🪛 SkillSpector (2.3.11)

[error] 11: [SC2] External Script Fetching: Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Remediation: Avoid downloading and executing remote scripts. Use trusted packages from PyPI/npm. If remote fetch is required, verify checksums and use HTTPS.

(Supply Chain (SC2))

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@plugins/superset/skills/superset/SKILL.md` at line 11, Replace the direct
curl-to-sh installation command in the Superset CLI instructions with a safer
installation method, preferably a package manager or signed release. If no
alternative exists, document downloading a pinned script or release, verifying
its integrity against a trusted checksum or signature, and only then executing
it.

Source: Linters/SAST tools


18-18: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Use the full CLI command in the core workflow.

tasks update <id-or-slug> omits the superset executable, so copying the documented step fails in a normal shell. Change it to superset tasks update <id-or-slug>.

🧰 Tools
🪛 SkillSpector (2.3.11)

[error] 11: [SC2] External Script Fetching: Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Remediation: Avoid downloading and executing remote scripts. Use trusted packages from PyPI/npm. If remote fetch is required, verify checksums and use HTTPS.

(Supply Chain (SC2))

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@plugins/superset/skills/superset/SKILL.md` at line 18, Update the “Plan work”
workflow instruction to use the complete `superset tasks update <id-or-slug>`
command, matching the existing `superset tasks create` command.

122-122: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Align the project/workspace prerequisites in the automation documentation.

Line 122 says a project ID is required to schedule automations, but the later workspace-only example permits omitting --project. Also, line 134 says every run produces a workspace, while line 137 says a new workspace is created only when --workspace is omitted. Document the project-scoped versus workspace-scoped behavior explicitly.

Also applies to: 134-137

🧰 Tools
🪛 SkillSpector (2.3.11)

[error] 11: [SC2] External Script Fetching: Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Remediation: Avoid downloading and executing remote scripts. Use trusted packages from PyPI/npm. If remote fetch is required, verify checksums and use HTTPS.

(Supply Chain (SC2))

🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@plugins/superset/skills/superset/SKILL.md` at line 122, Update the automation
documentation around the project ID prerequisite and the workspace examples to
distinguish project-scoped and workspace-scoped operations: clarify when
--project is required versus optional, and reconcile the statements about
whether runs always create a workspace or only create one when --workspace is
omitted. Ensure the documented behavior is consistent across the prerequisite
text and examples.

This branch was previously deployed

1 inactive deployment
preview — b27e072c Deployed Jul 24, 2026 by saddlepaddle via Deploy API #6823
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant