feat(demos): add visual recordings with Codex effort by harlan-zw · Pull Request #565 · skilld-dev/skilld.dev

harlan-zw · 2026-10-07T07:38:22Z

🔗 Linked issue

None.

❓ Type of change

📚 Description

Demo recordings need to identify the Agent, model, and effort that produced the output. Add eight visual demos from the weekly and monthly trending boards, and support sandboxed Codex recordings with an explicit model and effort. The Skill page Demo panel shows the recorded effort beside the model.

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

harlan-zw · 2026-10-07T07:50:18Z

Agent-authored recording evidence for head 7777b1855e0dc63cb524567c995e1c107d99c367.

Eight unique demos used Codex 0.160.1, gpt-6.1-sol, and medium effort. Latest-main duplicate check found zero overlap.

The full local suite passed 3,539 tests. Typecheck, lint, and build passed. All 22 media objects returned successful HEAD responses. The three films contain H.264 video and AAC audio, each 12.01 seconds.

Prism Pond, Fernhaven, Foxglove, and seed-library controls passed inside sandbox allow-scripts. Fernhaven's native-form failure was reproduced and repaired by the recording Agent. Its confirmation also passed through the actual demo route.

Game Builder's separate attempt returned five failed visual reviews and reached its runner limit. It is excluded from this pull request.

understand