perf(home): trim track card payloads

[harlan-zw]

🔗 Linked issue

Related to #610.

❓ Type of change

• 📖 Documentation
• 🐞 Bug fix
• 👌 Enhancement
• ✨ New feature
• 🧹 Chore
• ⚠️ Breaking change

📚 Description

The homepage carries 70 unused Skill examples and author totals in its track data. Serialize only the card fields used by the grid, preserving track order, links, counts, and avatars.

Replaying the captured homepage payload removes 7,443 bytes of inline data, about 6.6% of its 113,329 bytes. The API response stays unchanged.

🤖 AI disclosure: Harlan Agent Kit modified this description. My AI open-source policy.

[harlan-github-agent]

🤖 MERGED

Harlan Agent Kit posted this automated review. It is not Harlan's personal review or approval. AI open source policy. Last updated: 2026-10-10 11:46 UTC.

GitHub merged this pull request.

• Logged (10/100): Input type annotation for the transform drifts from the actual API response shape; harmless today but unverified against the endpoint type. View code Optional follow-up: Derive the transform input type from the /api/clusters handler's response type (or a shared type) instead of redeclaring it inline.
  • Ask an agent to verify and fix this finding

62f2a03c-537e-4773-96b3-183053bd311f

Selected findings run after merge and open separate pull requests.

[harlan-zw]

🤖 Harlan Agent Kit Agent verification of this change.

Captured production data contains 14 tracks and 70 unused Skill examples. Reserializing that captured homepage data without examples and author totals saves 7,443 bytes. This replay measures serialized data; it does not establish an indexing cause.

Local server-rendered data retains all 14 track records and removes both unused fields. Desktop at 1440px and mobile at 390px each render the same 12 demand-ranked cards, with titles, links, counts, and author avatars. Clicking the Design card loads its exact track heading. No page errors occurred during the checked navigation.

Local browser checks used polling after the host reached its file-watcher limit. The task-owned server and named pages were closed afterward.

[harlan-github-agent]

Harlan Agent Kit posted this automated review. AI open source policy.

Automated Review: READY.

Read the Review results