docs: add TweetClaw marketplace seed

[kriptoburak]

Summary

• add TweetClaw to the Search & Research seed data as a distinct OpenClaw plugin entry
• preserve explicit githubUrl and installCmd values when seeding skills
• make /api/agent Markdown use a stored install command when one exists

Validation

• npx eslint prisma/seed.ts src/app/api/agent/route.ts
• JSON seed-data integrity check: 5,148 skills, Search & Research count 343, TweetClaw entry present
• git diff --check
• added-line sensitive-value scan
• added-line dash scan

Full-repo checks still hit existing issues unrelated to this change:

• npm run lint fails in src/app/profile/page.tsx, src/app/skills/page.tsx, and src/app/trending/page.tsx for existing react-hooks/set-state-in-effect findings, plus existing warnings
• npx tsc --noEmit --pretty false fails on existing implicit-any and Prisma generated-client issues, including src/app/agents/[slug]/page.tsx, src/app/agents/page.tsx, src/app/api/user/skills/route.ts, and src/lib/prisma.ts
• npm run build compiles, then fails on the same existing TypeScript issue in src/app/agents/[slug]/page.tsx
• npm install --ignore-scripts completed but reported existing dependency audit findings; no dependency files are included in this PR

[oratis]

Review

Thanks for the contribution, and apologies for the long silence on this one. I checked the referenced repo (Xquik-dev/tweetclaw exists, active, 85★ as of today) — the project itself looks legitimate. A few issues with the approach, though:

1. The seed.ts update-path change would wipe enriched data on reseed

The update branch now writes githubUrl and installCmd unconditionally:

const githubUrl = skill.githubUrl || null;
const installCmd = skill.installCmd || `clawhub install ${skill.name || slug}`;

Almost none of the 5,147 existing seed entries carry a githubUrl, so re-running the seed against a live database would null out the real githubUrl values that our GitHub scraper has since populated, and overwrite scraper-enriched installCmd values with the generic fallback. That's a data regression — these two fields should stay create-only (or only be written on update when the seed entry explicitly provides them, with no fallback).

2. Vendor entries should go through the submit flow, not seed data

seed-data.json is a snapshot of the clawskills.sh dataset; hand-adding entries makes it diverge from its source and turns it into a listing channel. We have a normal submission path at https://takoapi.com/submit that goes through review — that's the right way to get TweetClaw listed, and it works without waiting on a release/reseed cycle.

Also, TweetClaw is an OpenClaw plugin (openclaw plugins install @xquik/tweetclaw) rather than a ClawHub skill, and the clawHubUrl (clawhub.ai/kriptoburak/xquik-tweetclaw) doesn't resolve to a skill page I can verify — the marketplace currently models ClawHub skills, so the entry would render with a misleading install command context.

3. The /api/agent change is good on its own

Preferring a stored installCmd over the generated clawhub install <slug> is a genuine improvement and independent of the listing. If you split that one-line change (plus the seed.ts fix per point 1) into its own PR, I'd be happy to take it.

Suggestion: close this in favor of (a) submitting TweetClaw via the submit flow, and (b) optionally a small PR with just the /api/agent install-command improvement.

[kriptoburak]

Rebased onto current main. The seed path now preserves per-skill githubUrl and installCmd values for any marketplace entry, the agent Markdown endpoint uses the stored install command, and the TweetClaw seed uses the verified npm install spec.\n\nValidation: targeted ESLint passes; TypeScript no-emit checking passes; seed JSON parses with exactly 5,148 skills and the Search & Research category count matches 343.