Sparki Codex Skill Setup and Troubleshooting
Install and connect the Sparki Codex plugin, choose the local or browser setup path, and recover an existing render when a download fails.

The Sparki Video Editor plugin has two setup paths. Codex desktop and local terminal sessions use sparki-cli. Hosted browser Codex uses a bundled Python runner. Choose the path that matches where the task runs.
Install from the current plugin repository
The Sparki Codex page provides a copyable installation prompt. The repository README contains the current marketplace commands. Ask Codex to follow that setup guide in your environment and check the result.
Connect from desktop or a local terminal
The local workflow uses uv to install the engine. If uv is missing, the skill asks before installing it through a trusted package manager. After plugin installation, the documented engine check and account connection are:
uv tool install --upgrade sparki-cli
uv tool run --from sparki-cli sparki --help
sparki connect --channel codex --timeout 540
If the direct sparki command is not available yet, replace its leading name with uv tool run --from sparki-cli sparki. That avoids depending on a shell restart. The connection command reuses a saved credential when available. Otherwise it opens Sparki authorization and waits for completion. Confirm connection success from the final successful result.
Connect from browser Codex
The current browser route requires Python 3 and uses the runner bundled with the plugin. Ask Codex to locate that runner from the installed skill, run its connectivity probe and return the authorization link. Complete the Sparki approval while the same authorization is pending.
The browser state is short-lived. The runner keeps the exchanged key in memory and removes the state after exchange. If the connectivity probe cannot reach agent-api.sparki.io, package installation will not restore that network connection. Resolve workspace access before creating the edit.
Review the editing task before it starts
Confirm which inputs are combined, the mode, aspect ratio, approximate duration and output location. Browser account authorization is followed by separate approval for the potentially paid editing project. Check Sparki’s current credit terms for the task you intend to run.
Recover a task that already exists
If browser rendering or delivery fails after the runner returned a task ID, keep that ID, mode and output path. Ask Codex to authorize again and resume that existing project. The documented resume operation polls and downloads the existing task.
Resume the existing Sparki project using its task ID, mode and output path. Recover the completed MP4 before considering another editing project.
For local CLI errors, ask Codex to inspect the installed command’s help and the error’s suggested action. Avoid repeating a project-creation request when you cannot tell whether it already created a task.
Handle account and storage limits
If credits are exhausted, review the account balance at Sparki before trying again. For full storage, review assets before choosing anything to delete. If a concurrency limit is reached, wait for an active project to finish. These messages describe different conditions and need different actions.
Find the output
The skill writes to ./sparki-output in the current working environment and reports the local path and output directory. Desktop sessions can reveal the file when a native file manager is available. Browser workspaces deliver the MP4 as an attachment when the host supports it. A path inside a hosted workspace needs file delivery before you can use the export on your own computer.
After setup succeeds, continue with the full editing walkthrough. Review the story, captions, sound and framing before publishing the returned video.
Setup and behavior checked against the Sparki plugin documentation and its current skill instructions.
https://sparki.io/blog/sparki-codex-skill-setup-troubleshooting
